Among 75 urosepsis isolates of Escherichia coli, 29 virulence factor (VF) genes were detected by use of a novel polymerase chain reaction (PCR) assay. Compared with probe hybridization, the PCR assay's specificity was 100% and sensitivity 97.1%. fyuA (yersiniabactin: overall prevalence, 93%), traT (serum resistance, 68%), and a pathogenicity-associated island marker (71%) occurred in most strains from both compromised and noncompromised hosts. Present in <20% of strains each were sfaS, focG (F1C fimbriae), afaldra, bmaE (M fimbriae), gafD (G fimbriae), cnf1, edtB (cytolethal distending toxin), cvaC (colicin V), and ibeA (invasion of brain endothelium). Different VFs were variously confined to virulence-associated phylogenetic group B2 (as defined by multilocus enzyme electrophoresis); concentrated in group B2, but with spread beyond; or concentrated outside of group B2. These findings provide novel insights into the VFs of extraintestinal pathogenic E. coli and demonstrate the new PCR assay's utility for molecular epidemiological studies.
Escherichia coli is the major cause of extraintestinal infections such as neonatal meningitis, gram-negative bacteremia, pyelonephritis, cystitis, and prostatitis [1–4]. Paradoxically, it also is the predominant facultative member of the normal human intestinal flora [5, 6]. Extraintestinal pathogenic E. coli and commensal E. coli typically differ with respect to phylogenetic background and virulence attributes. Pathogenic E. coli strains derive chiefly from phylogenetic group B2 (and to a lesser extent group D), as defined by multilocus enzyme electrophoresis [5, 7–11]. Commensal E. coli, in contrast, are characteristically from phylogenetic group A [5, 7, 8, 11]. Groups B2 and D comprise diverse evolutionary lineages that, because of their consistent associations with various extraintestinal infection syndromes, have come to be regarded as “virulent clones,” as traditionally defined based on O:K:H serotypes [9, 12–18].
A hallmark of such virulent E. coli clones is their possession of specialized virulence factors (VFs), traits that confer pathogenic potential and characteristically are infrequent among commensal strains [1, 2, 8, 19–21]. Recognized VFs of extra-intestinal E. coli include diverse adhesins (e.g., P fimbriae, S and F1C fimbriae, Drantigen specific adhesins, and type 1 fimbriae (which, unlike other VFs, are present in nearly all E. coli), toxins (e.g., hemolysin and cytotoxic necrotizing factor), siderophores (e.g., the aerobactin system), polysaccharide coatings (e.g., group II and group III capsules and lipopolysaccharide [LPS]), and invasins (e.g., IbeA, also called Ibe10) [19–23]. These VFs facilitate colonization and invasion of the host, avoidance or disruption of host defense mechanisms, injury to host tissues, and/or stimulation of a noxious host inflammatory response.
Informed selection of VFs to be targeted for prevention of extraintestinal E. coli infections (with their associated morbidity, mortality, and increased health care costs [24, 25]) requires knowledge of which VFs are prevalent in specific clinical syndromes and host populations, as revealed by epidemiological studies . Because host compromise can reduce the pathogenic importance of certain VFs [19, 27], it is also important to identify VFs that remain prevalent even among compromised hosts. Furthermore, because pathogenic behavior is predicted both by VF repertoire and by phylogenetic background [8, 15, 21, 28, 29], clonal associations of VFs must be evaluated.
During the past decade, new VFs have been described in E. coli [23, 30–35], and certain established VFs have attracted renewed attention [24, 36–45]. In addition, pathogenicity-associated islands (PAIs); that is, blocks of (known and suspected) VF genes that provide a mechanism for coordinate horizontal transfer of VF genes between lineages, and even between species, have emerged as a unifying characteristic of diverse pathogenic bacteria, including extraintestinal E. coli [21, 46–52]. The epidemiological studies called for by these recent discoveries should be facilitated by advances in polymerase chain reaction (PCR) technology and by the increasing availability of VF gene-sequence data, innovations that have permitted the development of PCR assays as an alternative to traditional DNA-probe hybridization methods for VF gene detection [36, 53–59]. In the present study, we developed and validated a multiplex PCR assay for 29 known and suspected VF gene regions of extraintestinal E. coli. We then used this new assay to define the prevalence, phylogenetic distribution, and associations with host compromise of these VF sequences among 75 well-characterized E. coli blood isolates from patients with urosepsis.
Primers' sequences (table 1) were as published, provided by other investigators, or selected de novo from available nucleotide sequences with the assistance of the application AMPLIFY (William Engels, University of Wisconsin Genetics Department, Madison, WI). For enhanced detection of partial copies of pap, primers specific for multiple regions within the pap operon were included, as were previously published primers specific for the 3 papG alleles . In addition to the central (consensus) region of the sfa/foc operon, as detected by published sfa primers , the adhesin genes of S fimbriae (sfaS) and F1C fimbriae (focG) were separately targeted. Dr family adhesins were detected by use of (consensus) afa/dra primers . Three other mannose-resistant adhesins of extraintestinal E. coli —that is, glucosaminyl-specific G fimbriae (gaf), M blood group antigen-specific M fimbriae (bma), and nonfimbrial adhesin-1 (nfa) —were also targeted, as was fimH, which encodes the mannose-specific adhesin subunit of type 1 fimbriae.
Toxins genes included were hlyA and cnf1, which have known associations with extraintestinal pathogenic E. coli , and cdtB (cytolethal distending toxin), which, to date, has been studied primarily in enteric E. coli and in genera other than Escherichia (the cdtB primers were designed and validated by Eric Oswald) [30, 31, 35, 61]. Siderophores included aerobactin (iutA) and yersiniabactin (fyuA), a Yersinia-associated siderophore system that recently has been identified along with other elements of the Yersinia-associated “high pathogenicity island” in some extraintestinal pathogenic E. coli strains . Capsule synthesis (kpsMT) primers intended to be specific for all group II, all group III, only K1, and only K5 capsule regions, respectively, were selected from aligned kpsMT sequences for K1 (group II), K5 (group II), and K54 (group III) capsules [22, 62, 63]. The rfc locus, which participates in O4 LPS biosynthesis , was targeted as a marker for (virulence-associated) E. coli serogroup O4 [2, 19, 65, 66]. cvaC, which encodes colicin V , was targeted as a marker for ColV plasmids, which have been proposed to confer enhanced virulence through their carriage of other specific VFs, including the aerobactin system and serum survival genes, such as traT and iss [19, 68–70]. The “invasion of brain endothelium” gene ibeA was included because of its association with neonatal meningitis and endothelial cell invasion [10, 23]. Finally, a coding region of unknown significance near the right-hand terminus of a sequenced PAI from archetypal uropathogenic strain CFT073 was used as a generic marker for uropathogenic PAIs .
Seventy-five blood culture isolates of E. coli collected in the mid-1980s from adults with urosepsis in Seattle, Washington, were studied. The status of these strains with respect to P fimbriae (phenotype, papEFG probe, and papG allele genotype), type 1 fimbriae (phenotype and fim probe), hemolysin (phenotype and papA probe), aerobactin (phenotype, iuc probe, and plasmid vs. chromosomal location of iuc), resistance to 12 antimicrobial agents, O:K:H serotype, and carboxylesterase B electrophoretic type (CBT), in comparison with host compromise status, have been reported elsewhere [27–29, 41, 71]. Strains were considered to derive from E. coli phylogenetic group B2 if they exhibited CBT B2 and from other phylogenetic groups if they exhibited CBT B1 [8, 72]. They were regarded as belonging to a particular O:K:H serotype if they exhibited 2 of the 3 corresponding antigens plus no other O, K, or H antigen . The antimicrobial resistance score was the number of agents to which the strain was found to be resistant . As additional controls for group III capsule synthesis genes (kpsMT III) and rfc, 14 strains of serogroup O4 that were previously reported as belonging (n = 11) or not belonging (n = 3) to a “J96-like” clonal group were studied [65, 66]. Other strains used during assay development as controls for specific VFs included J96 (papA, papC, papEF, papG alleles I and III, sfa/foc, focG, fimH, kpsMT III, hlyA, cnf1, and rfc) [65, 73], IA2 (papG allele II) , LG1315 (iutA) , P678-54/pHK1 1 (cvaC) , DH5a/pCIB10B (ibeA) , IH11165 (bmaE, gafD) [75, 76], HB101/pSR366 (kpsMT II) , P678-54/pAH1010 (nfaE) , E6468/62 (cdtB) , P678-54/pKT107 (traT) , 536-21/pANN801-13 (sfa/foc and sfaS) , NS24 (K1) , GR12 (K5) , and A30 (afa/dra; provided by Bogdan Nowicki). Strain P678-54 was used as a negative control . Strains were stored at −70°C in Luria broth plus 15% glycerol until ready for use.
Primers for each VF were first validated individually by use of template DNA from appropriate positive and negative control strains. Primers were then sorted into 5 pools (as listed below) according to primer compatibility and amplicon size, such that all desired products could be amplified in 5 separate PCR reactions and the products from each reaction could be resolved by size in gel electrophoresis (figure 1). Each primer pool was validated by use of pooled control DNA containing all relevant VFs. Alternative primers were selected and PCR conditions were adjusted as needed to yield simultaneous amplification of all desired products (figure 1).
Amplification was done in a 25-μ Lreaction mixture containing template DNA (2 μ Lof boiled lysate ), 4 mM MgCl2, 0.8 mM each of 4 dNTPs, 0.6 μ M of each primer (except for those marked “*,” which were used at a concentration of 0.3 μ M), and 2.5 units AmpliTaq Gold in 1× PCR buffer (Perkin Elmer, Branchburg, NJ). The 5 primer pools, with the 29 primer pairs listed in order of decreasing amplicon sizes (bp) within each pool, were as follows: pool 1: PAI (925), pap A (717), fimH (508), kpsMT III (392), papEF (326), and ibeA (171); pool 2, fyuA (787), bmaE (507), sfa/focDE (410), iutA (302), papG allele III (internal; 258), and K1 (153); pool 3: hlyA (1177), rfc (788), *nfaE (559), *papG allele I (internal; 464), *kpsMT II (272), and *papC (205); pool 4: gafD (952), cvaC (697), cdtB (430), focG (364), traT (290), and papG allele II (internal; 190); and pool 5: papG allele I (flanking; 1190), papG alleles II and III (flanking; 1070), *afa/draBC (594), *cnf1 (498), *sfaS (244), and K5 (159).
Reactions were heated to 95°C in an automated thermal cycler (PTC-100-96; MJ Research, Watertown, MA) for 12 min to activate the AmpliTaq Gold. This was followed by 25 cycles of de-naturation (94°C, 30 s), annealing (63°C, 30 s), and extension (68°C, 3 min) and a final extension (72°C, 10 min). Samples were electro-phoresed in 2% agarose gels, then stained with ethidium bromide, destained with distilled water, and photographed by use of an ultraviolet transilluminator and digital capture system (Gel Doc; Bio-Rad, Hercules, CA). The sizes of the amplicons were determined by comparing them with a 100-bp DNA ladder (Gibco/BRL, Gaithersburg, MD), which was run in multiple lanes on the same gel.
Duplicate lysates prepared for each test strain from separate bacterial colonies were amplified in separate PCR runs. Any discrepancies between the results of duplicate determinations for each VF were resolved by further investigation as needed.
Dot-blot hybridization was used to validate the PCR assay for 24 of the 29 constituent primer sets; that is, all primers but those for the 3 papG alleles that had been validated elsewhere , and the K1- and K5-specific primers which were predicted to generate probes that would hybridize nonspecifically with any group II kpsMT region, irrespective of K type. Probes were generated and labeled with digoxigenin by PCR by use of the same primers as used in the PCR assay (table 1), as described elsewhere . Blotting (under high stringency conditions) and detection were done in duplicate, as described elsewhere , by use of the same template DNA samples as those used in the PCR assay.
Validation of the PCR assay
Among the 75 urosepsis isolates, DNA probes for the 24 VF genes that were validated by blotting yielded 721 positive and 1,079 negative hybridization results. The PCR assay correctly identified 700 of the 721 blot-positive results (sensitivity, 97.1%), and all of the blot-negative results (specificity, 100%; accuracy, 98.9%). The PCR assay was 100% sensitive for all individual VFs except nfaE (sensitivity undefined; zero prevalence), papAH (57/59, 97%), papEF (57/58, 98%), fimH (73/75, 97%), cnf1 (11/12, 92%), and kpsMT II (47/62, 76%). Of note, all 3 strains that were blot-positive but PCR-negative for papAH or papEFwere negative for P fimbrial expression, and 2 contained only partial copies of pap by blot and PCR. After exclusion of kpsMT II, overall assay sensitivity was 99.1%, specificity 100%, and accuracy 99.7%. Of the 15 strains that were kpsMT II probe-positive but PCR-negative, 12 were serologically K2; these accounted for all but 1 of the K2 strains (figure 2). Thus, probe-versus-PCR discordance for kpsMT II was 92% sensitive and 80% specific for K2 capsule, and in combination with O6 antigen positivity was 91% sensitive and 98% specific for serotype O6:K2:H1.
The K1 and K5 primers were validated by comparison of PCR results with serologically determined K antigens. The K1 primers reacted with 20 of 21 K1-antigen-positive strains (sensitivity, 95%) and with one K53-positive strain (specificity, 98%; figure 2). The K5 primers, although identifying 7 of 8 K5-positive strains (sensitivity, 88%), also reacted with 17 non-K5 strains (specificity for K5, 75%), which included all but 1 of the 18 strains that were positive by PCR for group II kpsMT but not for K1 (figure 2).
The rfc and kpsMT III primers were further validated by comparison with dot-blot hybridization among 14 O4 strains, of which 10 were previously shown to exhibit group III capsules and/or to hybridize with a group III-specific kpsMT probe. All 14 O4 strains were confirmed as positive for rfc by both blot and PCR, as expected (PCR assay accuracy, 100%). The PCR assay also was 100% accurate in identifying the group III kpsMT probe-positive strains (not shown).
Comparisons of proportions between groups were evaluated by using Fisher's exact test. Because of multiple comparisons, P values between.01 and.05 were regarded as reflecting possible statistical significance, and P values ⩽.01 as reflecting statistical significance. Comparisons of the prevalence of different VFs were evaluated by use of McNemar's test , and comparisons of antimicrobial resistance scores were tested by use of the Mann-Whitney U test.
Prevalence of VFs
Among the 75 urosepsis isolates, the 29 virulence gene regions ranged in prevalence from 0% (nfaE) to 100% (fimH). Various pap elements, fyuA, iutA, kpsMT II, traT, and the PAI marker each occurred in over 50% of strains (figure 2, table 2). Of the siderophores studied, fyuA was more prevalent than aerobactin (iutA; 93% vs. 80%; P <.05 McNemar's test) and was the most prevalent VF detected other than fimH. Of the sialosyl-specific adhesins studied, focG was more prevalent than sfaS (P <, McNemar's test). Of the toxins studies, hly A (41%) was more prevalent than cnf1 (16%), which in turn was more prevalent than cdtB (8%; for both comparisons, P <.02, McNemar's test).
Phylogenetic distribution of VFs
The various VFs exhibited distinctive associations with CBT (carboxylesterase B types) B1 and B2, and with specific O:K:H serotypes, suggestive of divergent patterns of phylogenetic distribution (figure 2, table 2). These could be categorized broadly as (1) concentration within CBT B2, (2) concentration within CBT B1, and (3) equal distribution between CBTs B1 and B2. Among the CBT B2-associated VFs, 3 subgroups were apparent. The first subgroup included those VFs, that is, sfa/foc, sfaS, focG, cnf1, cdtB, rfc, and ibeA, that occurred essentially only in CBT B2 strains, were present in <40% of the CBT B2 strains, and were limited to specific subsets of these strains. sfa/foc was concentrated in serogroups O2, O4, O6, and O18 (P <.001 vs. other CBT B2 strains and vs. all strains). focG was largely limited to serotypes O2:K5, O6:K2, and O6:K5 (P =.002 vs. other CBT B2 strains; P <.001 vs. all other strains). cnf1 was concentrated in serotypes O2:K5/K7:H1, O4:K12, and O6:non-K2 (P <.001 vs. other B2 strains and vs. all other strains). cdtB was narrowly limited to the non-K1 O2 strains (P <.001 vs. other B2 strains and vs. all other strains). rfc was limited to strains of serogroup O4 (P <.001) ibeA occurred in the sole strain of serotype O1 8:K1: H7 (the serotype of the source strain for this invasin gene) and in 2 other K1-positive strains (figure 2).
A second subgroup of CBT B2-associated VFs included pap and hlyA, which were present in most (50%–85%) but not all of the CBT B2 strains, and in a substantial minority (20%–50%) of CBT B1 strains. pap was concentrated within the more prevalent serotypes of both the CBT B2 and B1 groups (figure 2). hlyA was concentrated, among the CBT B2 strains, specifically in serogroups O2, O4, O6, O12, and O18 (P <.001 vs. other B2 strains), and among the CBT B1 strains in serotypes O2: K5 and O25:K2:H2 (P <.001 vs. other B1 strains).
A third subgroup of CBT B2-associated VFs included kpsMT-II and the PAI marker, which were present in essentially all (i.e., >95%) of the CBT B2 strains, and also in many (30%–60%) of the CBT B1 strains. Although their high prevalence among CBT B2 strains precluded serotype-specific associations, among the CBT B1 strains kpsMT-II (blot) was specifically associated with K types K1, K2, K5, and K52 (P <.001), and the PAI marker was associated with serotype O2:K5:H4 and serogroup O157 (P <.001).
The second major pattern of phylogenetic distribution, exemplified by bmaE and possibly cvaC, was an association with CBT B1. bmaE was confined to CBT B1 strains and was concentrated in serotype O8:K27:H- (P =.002 vs. all other strains). When cvaC (which was 3 times as prevalent among CBT B1 strains as among CBT B2 strains) did occur in CBT B2 strains, it was associated with serotypes O1:K1:H7 and O2:K1:H7 (P <.001 vs. other B2 strains).
The third major pattern of phylogenetic distribution was of approximately equal prevalence among CBT B2 and B1 strains, whether low (<10%); for example, afa/draBC, or high (>160%); for example, fimH, iutA, fyuA, and traT (although fyuA did exhibit a trend associating it with CBT B2). afa/draBC, which was uncommon in either CBT, was concentrated (within CBT B2) in serogroup O75 (P =.049) The highly prevalent and broadly distributed fimH, iutA, fyuA, and traT did not exhibit serotype-specific associations.
Associations between VFs
The VFs displayed distinctive and complex associations with one another (figure 3, table 2). Although papG alleles II and III were both associated with sfaS, they otherwise exhibited nonoverlapping associations with other VFs. papG allele II was associated with kpsMT II, fyuA, and the PAI marker, whereas papG allele III was associated with sfa/foc, cnf1, and possibly K5 and ibeA (figure 3). sfa/focDE occurred almost exclusively in pap-positive strains and was associated with hlyA, cnf1, and the PAI marker (figure 3). sfaS and focG were both strongly associated with cnf1. However, whereas sfaS was strongly associated with ibeA and was negatively associated with iutA, focG was strongly associated with hlyA and the PAI marker and (borderline) negatively associated with traT. afa/draBC was not significantly associated with other VFs. bmaE was negatively associated with papG, group II capsules, and the PAI marker. gafD occurred in only 1 strain, which was also bmaE-positive (P =.05 bmaE vs. gafD).
kpsMT-II was positively associated with fyuA, hlyA, and the PAI marker and negatively associated with cvaC. Among the group II capsules, K1 was negatively associated with iutA and possibly also with hlyA, whereas K5 (which included most non-K1, non-K2 group II capsules) was associated positively with cnf1 and cdtB and negatively with cvaC.
fyuA was highly correlated with the PAI marker. iutA demonstrated a borderline negative association with ibeA. All cvaC- positive strains were positive for both iutA and traT, with the cvaC-traT association being statistically significant (figure 3). Finally, the PAI marker was highly correlated with all pap elements (except for papG allele III) and with sfalfoc, focG, kpsMT-II (PCR and blot), hlyA, and fyuA, whereas it was negatively correlated with bmaE (figure 3, table 2).
Plasmid aerobactin systems
Presence of a plasmid-associated aerobactin system, as previously detected in 16 (21%) of the 75 urosepsis isolates [27, 29], was significantly associated with bmaE (P =.002), cvaC (P <.001), and (P <.001) and traT (P =.002) In addition to its previously demonstrated negative associations with pap elements and hly, plasmid aerobactin also was negatively associated with sfalfoc (P = .008), focG (P =.03), kpsMT- II (blot; P <.001), K5 (P =.01), and the PAI marker (P =.004).
Several VFs were associated with decreased antibiotic resistance, including all pap elements except papG allele III (median resistance score in strains with vs. without VF, 0 vs. 4; P <.001) sfa/foc (0 vs. 1; P = .03), kpsMT II (PCR; 0 vs. 3.5; P =.006), kpsMT II (blot; 0 vs. 5; P < .001), fyuA (1 vs. 4; P =.02), hlyA (0 vs. 1; P =.004), and the PAI marker (0 vs. 3.5; P <.001). Both focG and rfc exhibited similar trends (not shown). In contrast, cvaC and bmaE were significantly associated with increased antibiotic resistance (median resistance score among strains with vs. without VF: cvaC, 2 vs. 0.5; P =.04 bmaE, 6 vs. 1; P =.003 afa exhibited a similar trend (4 vs. 1; P =.09)
Of the VFs newly detected in the present study (fim, pap, hly, and iutA having been reported elsewhere), only sfaS and bmaE exhibited significant associations with host compromise. Both were actually more prevalent among hosts with an upper urinary tract abnormality than among other hosts (sfaS: 2/8 vs. 1/67; P =.03; bmaE: 2/8 vs. 2/67; P =.05). Neither these nor any of the other VFs was significantly associated with presence or absence of urinary tract abnormalities in general, urinary tract instrumentation, a medical illness, or a combination of these compromising conditions (not shown). On the contrary, several of the more common VFs were substantially prevalent among strains from both compromised and noncompromised hosts, including fyuA (91% vs. 97%, respectively), group II kpsMT (by blot, 61% vs. 90%), traT (63% vs. 76%), the PAI marker (67% vs. 76%), and fimH (all strains; for all comparisons between compromised and noncompromised hosts, P >.10).
In this study, we developed and rigorously validated a novel multiplex PCR assay for 29 putative VF gene regions of extraintestinal E. coli. In the process, we defined the prevalence, clonal distribution, and associations with antimicrobial resistance and host compromise of the virulence genes included in the assay. Our findings provide novel insights into the molecular epidemiology of extraintestinal E. coli VFs and demonstrate that the multiplex PCR assay is highly accurate and informative.
The assay was extremely accurate overall in detecting specific VF genes, with the exception of group II kpsMT sequences. That there were no false-positive results in comparison with probe hybridization reflects the specificity of the primers for the VF regions of interest under the stringent amplification conditions used. That several probe-positive strains were not detected by the pap AH, papEF, fimH, and cnf1 primers is probably attributable either to the known polymorphic nature of these and other VF genes [21, 38, 60, 84–86] or to partial deletions, as appeared possible in the 3 strains that were probe-positive but PCR-negative for pap elements and phenotype-negative for P fimbrial expression. Such infrequent false-negative PCR results should be of little epidemiological significance.
In contrast, the group II kpsMT primers appreciably lacked sensitivity, particularly for K2 capsules, and the K5 primers detected almost all non-K2 group II capsular types. Informed selection of improved group II and K5 primers will require sequence data for more than the currently-available K1, K5, and K54 kpsMT variants [22, 62, 63]. Interestingly, discrepancies between probe and PCR for group II kpsMT proved to be informative, in that most discrepant strains were of capsular type K2. At the practical level, this could provide a genetic means of tentatively identifying strains as K2, particularly if they are known to be O6. This is relevant to the identification of serotype O6:K2:H1, the predominant serotype in the present population and that of archetypal uropathogenic strains CFT073 and AD110 [51, 80].
The present study provides novel epidemiological information pertinent to the development of anti-VF interventions. The high prevalence of fyuA found among E. coli urosepsis isolates from Seattle confirms previously reported findings involving bacteremia isolates from Germany, among which fyuA was significantly more prevalent than among fecal control strains . Taken together, these data suggest that fyuA is broadly characteristic of bacteremic E. coli. The high prevalence of fyuA observed in the present study among isolates from compromised, as well as from noncompromised, hosts indicate that fyuA should be a particularly useful target for an intervention if fyuA can be demonstrated to actually contribute to (as opposed to merely serve as a marker for) invasiveness. Similarly, traT also was found to be substantially prevalent among both compromised and noncompromised hosts. Comparisons with appropriate control strains, and experimental evaluation of traT as a VF, are needed .
The high prevalence of the PAI marker observed in the present study is consistent with recently reported findings involving other clinical isolates [50, 51] and suggests that as-yet undefined genes associated with this and related PAIs also may contribute to urovirulence and may potentially serve as targets for interventions. As with fyuA, the minimal impact of host compromise on the prevalence of the PAI marker bodes well for the utility in compromised hosts of interventions directed toward this or other PAI-associated elements.
Several VFs that previously have received minimal attention in the context of urosepsis were found to occur in only a small minority (<20%) of strains each; therefore, although these VFs may contribute to the pathogenesis of urosepsis in selected patients, they would not be ideal targets for anti-VF interventions for the prevention of urosepsis. foc, an sfa-related MR adhesins, which, unlike sfa, is not a hemagglutinin , was actually much more prevalent among sfalfoc-positive strains than was sfa, a situation that would have escaped attention had foc- and sfa- specific primers not been used. ibeA, which we postulated might promote bloodstream invasion during urosepsis because it promotes penetration of the blood-brain barrier during meningitis , did not exhibit the high prevalence observed among neonatal meningitis isolates . cdtB was limited to O2 strains, consistent with the O2 status of the single previously-reported cdt-positive strain from a patient with urosepsis . Although to date cdt has been regarded primarily as an enteric VF, the present study's findings suggest that cdt should also be investigated as a possible extraintestinal VF . bma and gaf occurred primarily in strains lacking pap, sfalfoc, and afa, and hence presumably were these strains' primary MR adhesins and could have played a contributory pathogenetic role. Studies of all these “low-prevalence” VFs in other extraintestinal syndromes are needed.
Because CBT B2 equates with (virulence associated) phylogenetic group B2 [8, 72], and O:K:H serotypes usually correspond with clonal groups [14, 88], the availability of CBT and O:K:H serotype data for this collection allowed an assessment of the phylogenetic distribution of the various VFs. Among the group B2-associated VFs, those that were present in only a small minority of B2 strains (e.g., sfalfoc, sfaS, focG, cdtB, and ibeA) are likely to have been recently acquired by the B2 phylogenetic group. Their confinement to the B2 group (with 1 exception each for sfa/foc, focG, and cnf1, which suggest horizontal transfer out of group B2) may be due either to their recent arrival in the B2 group or to barriers (whether from genetic or selection factors) to their horizontal movement into other groups. The overlapping but not completely concordant distribution of serotypes (e.g., for cnf1 vs. cdtB), and the sporadic appearance of several of these traits even within a single O:K:H serotype (figure 2), cannot be readily accounted for by strict vertical inheritance, thereby strongly suggesting that some of these traits have moved horizontally between lineages within the B2 group.
Those B2-associated VFs that were present in most but not all B2 strains (e.g., pap and hlyA) probably have been acquired by the B2 group at earlier points in its evolutionary history, with pap probably preceding hlyA. However, their incomplete penetration through the B2 group suggests that they did not enter E. coli prior to the differentiation of B2 strains from other E. coli. Thus, the substantial prevalence of both pap and hlyA outside of the B2 group is strongly suggestive of horizontal transfer from B2 into other lineages . It should be noted that the evidence of pap remnants in O75 strains (figure 2 and unpublished data) suggests that pap may once have been more widespread within group B2 than is apparent at present, hence may represent a pre-B2 acquisition. In contrast, those B2-associated VFs that were nearly universally prevalent among the B2 strains (e.g., kpsMT-II and the PAI marker), and also were prominent among non-B2 strains, could well have been acquired by E. coli prior to the branching off of the B2 group from an ancestral trunk that also served as the source for certain non-B2 lineages; for example, virulence-associated phylogenetic group D . Alternatively, they may represent a very early acquisition by the B2 group, with subsequent horizontal transfer into selected non-B2 lineages .
Those VFs (e.g., cvaC and bmaE) that were concentrated outside of group B2 may have entered E. coli initially in non-B2 lineages subsequent to the branching off of group B2 . However, cvaC (with its associated iutA) clearly has moved horizontally between non-B2 and B2 strains, or has independently entered group B2, to take up residence specifically within serotypes O1:K1:H7 and O2:K1:H7. Finally, several VFs (e.g., fyuA and fimH) were highly prevalent throughout the population, suggesting that they either entered E. coli early during its evolutionary history, hence are now present in all members of the species (which is likely the case for fimH) or, if more recently acquired (e.g., as part of a PAI from another species, which may be the case for fyuA), are highly horizontally mobile, are strongly selected for in the context of urosepsis, or are both.
Some inferences also could be drawn regarding the mode of horizontal transfer of VFs between lineages. Because both sfaS and focG were confined to strains also positive for the PAI marker, and because foc occurs as part of a PAI in strain CP9 , both VFs may have been PAI-linked in the present population, including when appearing in non-B2 strains (i.e.,focG in strain 2P1). In contrast, pap elements and hly, although usually associated with the PAI marker, also appeared in its absence, particularly when occurring in non-B2 strains, consistent with horizontal mobility out of group B2 not as part of a PAI, or as part of a PAI not containing the PAI marker included in our assay. Finally, the appearance of afa and cvaC in disparate lineages, unassociated with one another or with other VFs, would be consistent with the known occurrence of these sequences on plasmids [36, 69]. It is also likely that iutA was colicin V plasmid-associated among the O1:K1:H7 and O2:K1: H7 strains, because these 2 serotypes had completely concordant results for iutA and cvaC.
In summary, we have developed and rigorously validated a novel multiplex PCR assay capable of detecting a broad array of established and putative extraintestinal E. coli VF genes with a high degree of sensitivity and specificity. fyuA, traT, and a PAI marker were highly prevalent among urosepsis isolates, including among pap-negative strains from compromised hosts, and hence may constitute useful targets for preventive interventions. Although most VFs were concentrated in phylogenetic group B2, many were also prevalent in other strains. The new PCR assay should facilitate the studies that are needed to define the epidemiology and phylogenetic associations of traditional and newer VFs in extraintestinal infections due to E. coli.
Control strains were provided by Gabriele Blum-Oehler, J. M. de Ree, Betsy Foxman, Lynne Gilson, Jorg Hacker, Richard Hull, Sheila Hull, James Kaper, Kwang Sik Kim, Timo Korhonen, Joel Maslow, Barbara Minshew, Harry L.T. Mobley, Lisa Nolan, Thomas Russo, Soren Schubert, Richard Silver, Ann Stapleton, Peter Williams, and Lori Wright. Dave Prentiss prepared figure 1. Diana Owensby assisted with manuscript preparation.