Molecular Epidemiological and Phylogenetic Associations of Two Novel Putative Virulence Genes, iha and iroN_{E. coli}, among Escherichia coli Isolates from Patients with Urosepsis

Strains.The clinical collection studied comprised 67 well-characterized blood culture isolates of E. colicollected from adults with urosepsis in Seattle, Wash., in the mid-1980s (12). The status of these strains with respect to multiple characteristics has been reported previously (10-13, 15). (Eight strains from the 75-member collection which exhibited amplification fingerprints inconsistent with their putative O:K:H serotype and carboxylesterase B type in experiments not yet published were excluded from the present study, leaving 67 evaluable strains as studied herein.)

Molecular detection of iha and iroN_{E. coli}. iha and iroN_{E. coli} were each detected in duplicate both by PCR and by dot blot hybridization. Primers for iha, which were based on GenBank accession no.AF126104 (27), were 5′-CTGGCGGAGGCTCTGAGATCA-3′ (IHA-F; forward) and 5′-TCCTTAAGCTCCCGCGGCTGA-3′ (IHA-R; reverse) (827-bp product). Primers for iroN_{E. coli}, which were based on GenBank accession no. AF135597 (23), were 5′-AAGTCAAAGCAGGGGTTGCCCG-3′ (IRONEC-F; forward) and 5′-GACGCCGACATTAAGACGCAG-3′ (IRONEC-R; reverse) (665-bp product). Probes were generated and PCR labeled as previously described (14) using the same primers as for PCR-based gene detection. Amplification and blotting procedures were as previously described (14, 15). Appropriate positive and negative controls were included in each assay. Discrepancies between duplicate blot or PCR determinations were investigated with additional replicates as needed.

Phylogenetic analysis of PVF distribution.Strains were sorted according to carboxylesterase B electrophoretic type (i.e., B2 or B1) and then by O:K:H serotype within each carboxylesterase B type. Serotypes with two or more representatives in the study population were considered to exhibit homogeneity with respect to a particular PVF if the PVF was present in or absent from all representatives of the serotype or was present in all representatives but one (if three or more representatives total).

Statistical analyses.Comparisons of proportions were tested using Fisher's exact test. Comparisons of prevalence within the same population were tested using McNemar's test (6). Comparisons of antibiotic resistance scores and host compromise scores were tested using the Mann-Whitney U test. The criteria for the different categories of statistical significance follow: P > 0.10, indifferent association; P ≤ 0.10, borderline significant trend; P < 0.05, possible statistical significance, and P < 0.01, statistical significance. Associations of iha and iroN_{E. coli} with another PVF were considered to agree when they were in the same direction and exhibited P values of <0.10, to somewhat disagree (regardless of direction) when one exhibited aP value of <0.05 and the other exhibited a Pvalue of >0.10, and to strongly disagree when the associations were in opposite directions and both exhibited P values of <0.10.

Prevalence of iha and iroN_{E. coli}. iha was detected in 37 (55%) and iroN_{E. coli} in 26 (39%) of the 67 urosepsis isolates (0.05 < P < 0.10, McNemar's test) (Table1). Probe and PCR yielded identical results for each gene (not shown). Overall,iha and iroN_{E. coli} were indifferently associated with one another (P > 0.10).

Associations with other PVFs. iha and iroN_{E. coli} each exhibited positive, negative, and indifferent associations with diverse other PVFs (Table2). Consistent with the occurrence ofiha in archetypal uropathogenic strain CFT073 on a PAI that also contains papG allele II and hly (8, 15), iha was strongly associated with papGallele II (and all other pap elements except papGallele III) and with hlyA (Table 2). Similarly, consistent with the discovery of iroN_{E. coli} in archetypal uropathogenic strain CP9 on a PAI that also contains papGallele III, hly, cnf, and foc,iroN_{E. coli} was positively associated with each of these traits, despite not being significantly associated with otherpap elements (Table 2).

However, associations of iha and iroN_{E. coli} with other genes from their respective PAIs in strains CFT073 and CP9 were not categorical. For example, ihaexhibited only a marginal statistical association with malX, the PAI marker used in the present study (Table 2), even though in strain CFT073 malX occurs on the same PAI as ihaand is separated from iha by <2 kb (8).iha occurred in seven strains that lacked the PAI marker, whereas the PAI marker occurred in 19 strains that lackediha (Table 1). Similarly, despite its association withpapG allele III, iroN_{E. coli} occurred in four pap-negative strains and in 19 strains that contained papG allele II rather than papG allele III, whereas papG allele III occurred in one strain withoutiroN_{E. coli} (Tables 1 and 2). An analogous discordance was also noted for both iha andiroN_{E. coli} with respect to hly(Tables 1 and 2).

Not only were iha and iroN_{E. coli}inconsistently associated with PVFs from their respective PAIs of origin, but they exhibited divergent associations with other PVFs.iha and iroN_{E. coli} were concordantly positively associated only with focG andhlyA, and exhibited no concordant negative associations (Table 3). In contrast, they exhibited discordant associations with 20 of the remaining PVFs (Table4). (Of note, whereas iha was strongly negatively associated with plasmid iut[aerobactin] and cvaC [colicin V], iroN_{E. coli} was strongly positively associated with plasmid aerobactin and exhibited a trend toward an association withcvaC [Tables 2 and 3].) Thus, iha andiroN_{E. coli} differed more often than they agreed with respect to associations with other PVF genes (Table 3).

Associations with papA alleles. iha andiroN_{E. coli} differed substantially with respect to their associations with individual papA alleles (Table4). iha was significantly associated with the F7-2 allele, which is present in the PAI of strain CFT073 that containsiha, and with the F10 and F16 alleles, whereas it was significantly negatively associated with the F11 allele (Table 4). In contrast, iroN_{E. coli} was positively associated with the F14 allele (P = 0.029), which is present in the iroN_{E. coli}-containing PAI from strain CP9.iha and iroN_{E. coli} exhibited (nonsignificant) opposing associative trends with all the remainingpapA alleles except F13 (Table 4).

Phylogenetic distribution of iha and iroN_{E. coli}.Although iha and iroN_{E. coli} were both somewhat more prevalent among carboxylesterase B2 strains, they exhibited divergent patterns of distribution with respect to individual O:K:H serotypes (Table 1). In contrast to the more homogeneously distributed iha,iroN_{E. coli} was sporadically distributed, occurring in many different serotypes but rarely in all members of any one serotype (Table 1). Although both PVFs occurred in a similar proportion of the 14 most prevalent O:K:H serotypes (i.e., 10 of 14 foriha versus 9 of 14 for iroN_{E. coli}),iha satisfied criteria for homogeneous distribution (as defined above [phylogenetic analysis]) in all 14 serotypes, as compared with only seven for iroN_{E. coli}(P < 0.02, McNemar's test) (Table 1).

Among the B2 strains, iha was almost universally prevalent except among O1:K1:H7 and O2:K1:H7 strains, from which it was notably absent (P < 0.001 versus other B2 strains). In contrast, although iroN_{E. coli} was also absent from the five O1:K1:H7 strains, it occurred in three of the five (iha-negative) O2:K1:H7 strains and was absent from all four (iha-positive) O75 strains (Table 1). Likewise, among the B1 strains, whereas iha was largely confined to serotypes O7:K1, O15:K52, O25:K2, and serogroup O157 (P < 0.001 versus other B1 strains), iroN_{E. coli} was confined to strains of serogroups O8 and O9 (P < 0.001 versus other B1 strains), which resulted in a significant negative association of iha with iroN_{E. coli}among B1 strains (P = 0.004) (Table 1).

Host compromise and antimicrobial resistance. ihaand iroN_{E. coli} differed with respect to their associations (or lack thereof) with host compromise and antimicrobial resistance. iha was consistently more prevalent among strains from noncompromised hosts than those from compromised hosts, with the strength of the association depending on the type of host compromise (Table 5). In contrast, iroN_{E. coli} was more prevalent among strains from compromised hosts (Table 5). Host compromise scores were significantly lower among iha-positive strains (mean, 0.8 [range, 0–2]) than among iha-negative strains (mean, 1.4 [range, 0–3]; P = 0.002) but did not differ betweeniroN_{E. coli}-positive and -negative strains (means of 1.0 and 1.1, respectively [range for both groups, 0–3]; P > 0.10). Among strains from noncompromised hosts,iha was always somewhat or significantly more prevalent than was iroN_{E. coli}, whereas among strains from compromised hosts, iha and iroN_{E. coli} were similarly prevalent (Table 5).

Resistance to one or more antimicrobial agent was less common amongiha-positive strains than among iha-negative strains (38 versus 67%; P = 0.03), whereas it was comparably prevalent among iroN_{E. coli}-positive and -negative strains (46 versus 54%; P > 0.10). Similarly, antimicrobial resistance scores were lower amongiha-positive strains than among iha-negative strains (mean of 0.8 [range, 0–7] versus 3.0 [range, 0–10]; P = 0.002) but were similar among iroN_{E. coli}-positive and -negative strains (mean of 2.0 [range, 0–10] versus 1.7 [range, 0–9]; P > 0.10).

Implications of findings.That iha andiroN_{E. coli} were both significantly associated with other PVF genes from their respective PAIs in archetypal ExPEC strains CFT073 (iha) and CP9 (iroN_{E. coli}) suggests, as a first approximation, thatiha and iroN_{E. coli} occur consistently on the same (or similar) PAIs throughout the ExPEC population. Consequently, the striking differences documented in the present study between iha and iroN_{E. coli} with respect to their individual associations with other bacterial and host characteristics demonstrate that such associations are likely to be specific to individual PAIs and/or PVF genes rather than being common to all PAIs or to PAI-linked PVF genes in general. This implies that discovery that a particular PVF of ExPEC is linked to a PAI is insufficient to predict its associations with other bacterial and host characteristics, which instead can be discovered only through specific study of the particular PVF itself or the individual PAI.

However, our findings also suggest that the concept of an individual PAI as a discrete entity which can be tracked through the ExPEC population may itself be illusory (1, 2, 7, 8, 23, 26). The associations observed in the present study of iha andiroN_{E. coli} with other PVFs from their respective PAIs in source strains CFT073 and CP9, although statistically significant, were not absolute, evidence which suggests that the corresponding PAIs may have undergone considerable recombination over time, with acquisition or deletion of various PVF genes, or that each PVF has independently entered multiple different PAIs. Specific evidence of the fluidity of PAIs was found in the discordant occurrence of iha and the PAI markermalX (Tables 1 and 2), genes which on their shared PAI in strain CFT073 are located within 2 kb of one another (8). Thus, it is likely that at the same time that PAIs are generating diversity within the species by introducing new genetic material into disparate lineages via horizontal transfer (7), PAIs themselves are undergoing diversification through deletion and insertion of PVF genes, some of which may be more mobile (or labile) than are the PAIs themselves. PAIs thus may exist as continuously evolving mosaic constructs, a phenomenon which would be predicted to confound efforts to define a reliable marker for the PAIs of ExPEC in general or for any particular PAI. An example of this phenomenon may be the apparent insertion of the group III capsule genes into the group II capsule locus in strain CP9 (24). Similarly, TAI, the PAI on which iha was first identified, was conserved in structure in some, but not all, diarrheagenic E. coli tested (27).

The respective patterns of phylogenetic distribution of ihaand iroN_{E. coli} suggest divergent evolutionary histories and mechanisms of horizontal mobility, iha was present in almost all strains of carboxylesterase B type B2 except for those of serotypes O1:K1:H7 and O2:K1:H7 and was also highly prevalent in selected lineages among the non-B2 strains. This is consistent with early acquisition of iha by the B2 phylogenetic group soon after the differentiation of B2 strains from other E. coli(18) but after the O1:K1:H7 and O2:K1:H7 clonal group (which constitutes a discrete evolutionary cluster within the B2 group [19]) had split off from a common B2 ancestor. In this scenario, the appearance of iha among non-B2 strains would require the independent acquisition of iha by non-B2 lineages through horizontal transfer either from a B2 source or from an unknown external source. As an alternative scenario, ihacould have entered E. coli prior to the differentiation of the B2 phylogenetic group and subsequently have been deleted from the ancestor of the O1:K1:H7 and O2:K1:H7 branch of the B2 group after the differentiation of this branch from other B2 strains. In this scenario,iha could have been inherited vertically from a common ancestor by both B2 and certain non-B2 strains, for example, those of closely related phylogenetic group D (18).

In contrast to iha, the sporadic appearance ofiroN_{E. coli} in many different serotypes strongly suggests either multiple horizontal acquisition events or multiple scattered deletions of iroN_{E. coli} within otherwise iroN_{E. coli}-positive lineages. Although in strain CP9 iroN_{E. coli} occurs on a genomic PAI (23), the possibility of plasmid-associated transfer of iroN_{E. coli} in some strains was suggested in the present study by the striking statistical association of iroN_{E. coli} with cvaC (colicin V), which occurs on large conjugative plasmids (28-30). Precedent in ExPEC for a siderophore system which occurs variously on colicin V plasmids and on the genome is provided by the aerobactin system, which has been postulated to represent an extinct transposon (22, 28). The finding that in strain CP9 iroN_{E. coli} is flanked by IS1230sequences (23) suggests the possibility of independent mobility of iroN_{E. coli}, which would be consistent with exchange of iroN_{E. coli} between genome and plasmids.

Although direct genetic linkage is a plausible explanation for the observed statistical associations between different putative virulence genes, other explanations should be considered. Alternative mechanisms include coselection for the associated traits (independent of these traits' genetic relationship to one another), and coselection for unmeasured traits that are genetically linked to the statistically associated putative virulence genes. Confirmation of actual genetic linkage between putative virulence genes would require direct genetic analyses. The results of the present study can assist the design of such analyses by suggesting which linkages should be sought experimentally.

The significantly higher prevalence of iha found in the present study among E. coli bacteremia isolates from noncompromised hosts than in those from compromised hosts suggests thatiha (or VFs linked to iha) may assist in overcoming host defenses that are breached in patients with the types of compromise studied (12). In contrast, our observation that the prevalence of iroN_{E. coli} increased in the presence of host compromise suggests that unlike iha,iroN_{E. coli} should be as good a target for an anti-VF intervention among compromised hosts as among noncompromised hosts. The absence of an association of iroN_{E. coli} with antimicrobial resistance suggests that iroN_{E. coli} should also remain as effective a target for an anti-VF intervention among multidrug-resistant E. coli as among susceptible strains, whereas iha, which was negatively associated with antimicrobial resistance, would be most useful as a target among antimicrobial-susceptible strains.

It should be noted that the present study population represented a single geographical locale, a single clinical syndrome, and a time period of several years in the mid-1980s. Analysis of strains from other locales, clinical syndromes, and time periods for ihaand iroN_{E. coli} would provide a valuable complement to the present study.