Sequencing and Functional Annotation of Avian Pathogenic Escherichia coli Serogroup O78 Strains Reveal the Evolution of E. coli Lineages Pathogenic for Poultry via Distinct Mechanisms

Comparison of the genome sequences of ST23 serogroup O78 strains with that of the APEC ST95 serogroup O1 strain.Surveys of the prevalence of specific genes have indicated that APEC strains are highly heterogenous; however, the full extent of inter- and intraserogroup and sequence type diversity across genomes is ill defined. We therefore sequenced and annotated the chromosomes and plasmids of χ7122 and IMT2125, representing the prevalent ST23 serogroup O78 lineage, as described in Materials and Methods. Circular diagrams showing features of the chromosomes of the two strains were made (see Fig. S1 and S2 in the supplemental material). The sequences of the four plasmids of χ7122 have been described elsewhere (43, 44). Four contiguous plasmids of IMT2125 were also obtained. Our analysis of the APEC O1 sequence annotated by Johnson et al. (4) indicated, on the basis of conservation with the latest E. coli K-12 MG1655 annotation, that some gene predictions and start sites are erroneous. To ensure that this did not suggest a higher number of strain-specific features, we ran the same annotation pipeline as that used for the APEC O78 strains on the raw APEC O1 sequence.

For comparison, we also obtained the draft sequence of an ST23 serogroup O78 human ETEC strain (E85). We used the consistent annotation of the genomes to identify 3,153 shared core proteins. RAxML was then used to examine the phylogenetic relationships between the ST23 and ST95 strains, as well as between these strains and reference genomes in public databases for eight other E. coli strains representing different pathotypes. The analysis supported the clustering of APEC ST95 O1 with uropathogenic E. coli reported previously but clearly indicated that APEC ST23 O78 strains are distinctly and distantly clustered (Fig. 1). Indeed, the APEC O78 strains occupy positions on the tree that suggest they are closely related to the ETEC strain of the same serogroup and sequence type, even though strain E85 is associated with acute enteritis in humans rather than with systemic avian disease. These findings challenge the prevailing view that APEC strains are closely related to other ExPEC strains, and to UPEC strains in particular. In turn, the data suggest that broad classification of E. coli strains into diarrheagenic and extraintestinal categories is not always consistent with the phylogenetic signatures present in the chromosomal sequences of the strains.

• Open in new tab
• Download powerpoint

Fig 1

Phylogenetic tree generated using RAxML (version 7.0.4) from the concatenated sequences of the 3,153 proteins that are shared by the 12 E. coli strains analyzed (see Materials and Methods). Pathotypes are color-coded according to the key, and serotypes are given at the position of each strain on the tree.

Given the clustering of APEC ST23 strains and their phylogenetic distance from the APEC ST95 strains, we next examined the extent of conservation of genes within and between the sequence types in order to assess the extent to which they are adapted to cause the same disease via acquisition of common or unique determinants. In general, pairwise analysis of the chromosome sequences of χ7122, IMT2125, and APEC O1 revealed conservation of the sequence and order of genes (see Fig. S3 in the supplemental material). The genome sequences obtained for χ7122, IMT2125, and APEC O1 are predicted to encode 4,627, 4,638, and 5,057 genes, respectively. OrthoMCL analysis was used to search for shared and strain-specific orthologous proteins. This analysis identified 3,782 chromosomally encoded core predicted proteins shared by all three strains, but it also identified predicted proteins unique to χ7122 (n, 288), IMT2125 (n, 298), and APEC O1 (n, 1,154) (Fig. 2). A relatively small number of orthologous proteins were shared only by APEC O1 and χ7122 (n, 60), or only by APEC O1 and IMT2125 (n, 61), whereas the two APEC O78 chromosomes encoded 497 orthologous proteins that were not shared by APEC O1. Taken together, these data indicate marked divergence of the O78 strains from the sequenced APEC O1 strain but a relatively high degree of similarity of the O78 strains to each other.

• Open in new tab
• Download powerpoint

Fig 2

Venn diagram showing the numbers of conserved and strain-specific orthologous proteins predicted by OrthoMCL to be encoded by the chromosomes of strains χ7122, IMT2125, and APEC O1.

Because the salient features of the APEC O1 genome have been described previously (4), emphasis is placed here on further differences within and between serogroups and sequence types. A striking feature of this comparison is the low extent of conservation in χ7122 and IMT2125 of the 43 genomic islands of APEC O1 described by Johnson et al. (4). Thirty of the 43 genomic islands of APEC O1 were absent or poorly conserved in the two APEC O78 strains. These 30 genomic islands account for the bulk of the APEC O1-specific orthologs described above and include 4 predicted pathogenicity islands (PAIs) (PAI I_{APEC O1} [∼83 kb] near tRNA-Phe, carrying tia, ireA, and pap; PAI II_{APEC O1} [∼28 kb] near tRNA-Asp; PAI III_{APEC O1} [∼19 kb] near tRNA-Thr, carrying vat; PAI IV_{APEC O1} [∼87 kb] near tRNA-Asn, encoding yersiniabactin). The lack of conservation of the 43 genomic islands described previously for APEC O1 in χ7122 and IMT2125 is consistent with the variable conservation of 12 such islands among other APEC isolates in a recent survey (9). Other virulence-associated loci present in APEC O1 but absent from χ7122 and IMT2125 include the chu operon, involved in iron uptake; the kps locus, mediating group II capsule synthesis; and genes encoding cytolethal distending toxin, Auf fimbriae, and Ibe invasion-associated proteins. The sitABCD genes in APEC O1 island 10 were absent from this position in χ7122 but are harbored by pChi7122-1 in this strain (44). Prophages exhibited partial homology between the strains and accounted for some of the differences between the O78 strains. Some prophages were conserved in APEC O1 and one O78 strain but not the other (e.g., those between yfdC-dsdA and ydfI-rspB are mostly shared by APEC O1 and χ7122). As expected, genes associated with lipopolysaccharide core biosynthesis were conserved in all three strains but not in E. coli K-12, whereas differences exist between APEC O1 and the serogroup O78 strains in the region between gnd and galF that determines O-antigen biosynthesis. The few islands that were conserved across all three strains (e.g., islands 18, 26, and 37) were also conserved in other sequenced ExPEC strains (4, 9).

Many of the orthologs specific to the APEC ST23 O78 strains are encoded in clusters (see Fig. S1 and S2 in the supplemental material). Among the virulence-associated loci present in χ7122 and IMT2125 but absent from APEC O1 is the degenerate type III secretion locus ETT2, located between yqeG and the glyU tRNA locus in many E. coli strains (63). The ETT2 locus in χ7122 and IMT2125 is truncated from within eivA to the end of the locus described for E. coli O157:H7 strain Sakai, consistent with the truncation described for another sepsis-associated APEC O78 strain, termed 789 (30). Within ETT2, the predicted spaS and eivJ genes contain distinct frameshift mutations in χ7122 and IMT2125, consistent with the varied patterns of mutational attrition of the locus reported among E. coli strains (30, 63). The ymcDCBA-yccZ-etp-etk genes, which determine O-antigen capsule biosynthesis (located between appA and yccM) (64), were conserved in both χ7122 and IMT2125 but absent in APEC O1. The same applies to the stg fimbrial locus (located between pstS and glmS), which influences air sac colonization in chickens (29), and two cryptic fimbrial loci (the sfm locus, between folD and intD, and the yraHIJK genes, between agaI and yraL). Forced induction of these operons has established that they encode functional adhesins in E. coli K-12 (65). A ca. 62-kb region between yjgB and yjhR, carrying the fecIRABCDE genes for citrate-dependent iron(III) uptake, genes encoding a putative sugar uptake phosphotransferase system (PTS), and numerous genes of unknown function, was also largely conserved in χ7122 and IMT2125 but absent in APEC O1.

Many of the APEC ST23-specific functions are putatively related to metabolism and include clusters of genes predicted to mediate 4-hydroxyphenylacetate degradation (between ygiY and tsr), cyanate metabolism (between codA and lacA), use of 3-(3-hydroxyphenyl)propionic acid (mhp operon, between lacI and yaiL), the dioxygenolytic pathway for initial catabolism of 3-phenylpropionic acid (hcaRACBD, between hcaT and yphA), and fructoselysine catabolism (yhfLMNOPQR, between cysD and yhfS), as well as a putative 10-subunit hydrogenase complex (hydrogenase-4) encoded by the hyf locus (hyfABCDEFGHIJ-hyfR-foc, between bcp and uraA). Other metabolism-related clusters of less-defined function encoded by χ7122 and IMT2125 but not APEC O1 were also detected (e.g., between ymjA and pspF, ydbL and ynbD, hipB and uxaB, yeaS and rnd, and yqiN and parE).

Comparison of the genome sequences of APEC serogroup O1 and O78 strains extends observations made by Brzuszkiewicz et al. in relation to the evolution of virulence in another group of extraintestinal E. coli strains (66). These authors reported that two human uropathogenic E. coli (UPEC) serogroup O6 strains differed markedly in the content of the accessory genome, despite the fact that they were adapted to cause the same disease. Approximately 430 genes specific to each UPEC O6 strain were identified, and these were distributed mostly in pathogenicity islands and smaller islets (66).

In vivo functional annotation of the genome of APEC ST23 O78 strain χ7122.Since analysis of genome sequences and phylogeny indicated the evolution of APEC from distinct lineages, and since preceding studies mostly used ST95 strains, we sought to identify virulence factors of APEC ST23. A library of 2,185 signature-tagged mini-Tn5Km2 mutants of strain χ7122 was screened in turkeys for mutants defective in colonization of the respiratory tract and systemic translocation. Pilot studies indicated that strain χ7122 produces airsacculitis, perihepatitis, and pericarditis typical of natural infections when administered to the left caudal air sac of a 3-week-old turkey poult, with recovery of ca. 10⁷ CFU/g from the lung and liver 24 h postinoculation. In contrast, a Δrfb O-antigen-deficient mutant of χ7122 was recovered at ca. 10² CFU/g from these sites 24 h after the inoculation of age-matched animals in the absence of overt pathology, in agreement with published data (16). It was not feasible to extend the time interval postinoculation beyond 24 h, because most birds infected with χ7122 presented symptoms consistent with agreed humane endpoints; however, it was evident from these pilot studies that adequate time had elapsed for attenuation to be evident. Inoculation of three separate birds with a single pool of 95 signature-tagged mutants revealed that the same mutants were reliably negatively selected, with minimal variance in the relative intensity of hybridization signals between replicates and no evidence of stochastic loss of mutants in lung or liver homogenates (which would be indicative of bottleneck effects) (data not shown). We therefore screened the entire library in pools of 95 mutants, giving each pool via the intra-air sac route to three 3-week-old turkey poults, with recovery of mutants from lung and liver homogenates 24 h postinoculation.

The composition of input and output pools was examined by amplification and detection of signature tags, and a total of 119 mutants were inferred to be attenuated on the basis of their absence or negative selection in the output pool relative to the inoculum. We determined the insertion sites in 75 such mutants by subcloning and sequencing (Table 1). Conservation of the mutated genes in IMT2125 and APEC O1 is indicated by footnotes in Table 1. Multiple attenuating mutations were independently detected in several genes (evgS, exoP, intQ, 5′ of traM, traM-MM2_058, trbB, yadL, ycbY, ycgK, ygaV, yjiZ, ynbB), including a number of siblings with the same tagged transposon and insertion site that were negatively selected in separate pools (evgS, intQ, 5′ of traM, trbB, ycbY, ycgK, ygaV, yjiZ, ynbB). The reliable negative selection of such mutants provides a measure of confidence that mutants are absent owing to their genotype rather than to stochastic loss due to bottleneck effects or to sampling populations of inadequate size. This is further supported by the detection of multiple attenuating mutations in the same pathway. For example, five mutations were identified in the yadN-ecpD-htrE-yadMLKC cluster, predicted to encode fimbrial subunits and a chaperone-usher secretion system homologous to those involved in the production of E. coli type I fimbriae. A total of 10 mutations predicted to impair F-pilus assembly and conjugal transfer were identified, with most mapping to pChi7122-2 (5′ of traM, trbB, trbF, traM-MM2_058) and single insertions in traE on pChi7122-1 and traJ on pChi7122-3.

A number of attenuating mutations detected in the present screen support observations from other studies on APEC pathogenesis in poultry. Thus, a flagellin subunit (fliC) mutant of χ7122 was not recovered from the lung or liver (Table 1), in agreement with the finding that fliC mutation in APEC O78 strain EC34195 impaired intestinal persistence in 10-day-old chicks and systemic virulence in day-old chicks after oral dosing (28). A signature-tagged χ7122 mutant with an insertion in the epaP gene, located in the ETT2 locus, was negatively selected in the liver, a finding consistent with the reduced lethality of a null mutant of APEC O78 strain 789 lacking the putative inner membrane ring of the secretion complex (ΔeprHIJK) after intraperitoneal infection of day-old chicks (30). In common with the present survey, an attenuating mutation in the putative polysaccharide hydrolase YcjM was detected by STM analysis of random APEC O2 mutants 48 to 72 h after intratracheal inoculation of chickens (15). Similarly, attenuating mutations detected in yahB and ykgB in the present screen are supported by the isolation of attenuating mutations in the adjacent genes (yahA and ykgC) in APEC O2 (15). Screening of the same APEC O2 library in a model of early lung colonization identified attenuating mutations in evgA and tdcA (21), consistent with the isolation of evgS and tdcC mutants here.

A mutant with an insertion in ybcU was negatively selected in the lung and liver and is predicted to lack a protein with 91% amino acid identity over 97 residues to Iss, encoded on pChi7122-1 (44), and 93% identity to the Bor protein of APEC O1 and lambdoid phages. Both Iss and Bor are localized on the APEC surface (67) and influence survival in serum when cloned into E. coli K-12 (68, 69). A further closely related homolog of APEC O1 Iss and Bor is encoded by orf01374 in χ7122; however, we were able to fine map the attenuating insertion to ybcU, since the insertion is proximal to the 5′ end of ybcU, and the upstream sequence captured in the transposon (Tn)-flanking region is distinct from that upstream of iss on pChi7122-1 and from that of the chromosomal orf01374 gene. Gene prevalence surveys have associated iss carriage with APEC virulence; however, a defined iss deletion in pAPEC-O2-ColV did not impair lethality for chick embryos despite induction of the gene in vivo (70), and the iss gene carried on pChi7122-1 does not play a major role in the serum resistance of χ7122 (16). Isogenic iss and bor APEC mutants do not exhibit the same levels of serum sensitivity (71), and the relative contributions of this family of proteins to strain χ7122 pathogenesis merit further study.

The apparent role of plasmid-encoded conjugal transfer systems in APEC pathogenesis is supported by other studies. For example, the APEC O1 traY and traQ genes are upregulated in chicken serum (72), and traJ has been reported to be induced during invasion of human brain microvascular endothelial cells (HBMECs) by neonatal meningitis-associated E. coli K1 (73). Moreover, traJ mutants show impairments in the invasion of HBMECs in vitro (73, 74), the penetration of the blood-brain barrier after intracardiac inoculation of neonatal rats (74), and dissemination from mesenteric lymph nodes to the blood and visceral organs after oral dosing (75).

No attenuating mutations were detected in several known virulence-associated loci of APEC, including genes mediating lipopolysaccharide biosynthesis or the production of type I, Stg, or Yqi fimbriae, and the iro, iuc, and sit genes, associated with iron uptake. It is not clear whether this reflects the absence of such mutants from the library or features of the infection model. Similarly, the genes identified in the present screen differ from those implicated in NMEC and UPEC virulence in rodent models (76–78). A distinct set of virulence factors were identified by STM when random APEC O2 mutants were screened 48 to 72 h after intratracheal infection of chickens (15) compared to an early lung colonization model (20, 21). None of the attenuating mutations detected by STM in χ7122 occurred in APEC O78 genes reported to be upregulated in chickens in a recent promoter trap screen (79). Differences between surveys reinforce the parallel evolution of avian pathogenic strains with distinct sets of virulence-associated genes and may also reflect differences in the host, inoculation route, and dose.

Validation of mutant phenotypes.To confirm the attenuation of mutants with defects in specific loci, λRed mutagenesis was used to replace six genes identified in the primary screen (cadA, evgS, fliC, fucP, htrA, yccC [etk]) with a kanamycin resistance cassette. Each mutant, as well as the parent strain, was then administered separately to the left caudal air sacs of three 3-week-old turkey poults, and bacteria in the lung and liver 24 h postinoculation were enumerated. Seven transposon mutants from the primary screen (with insertions in entF, exoP, traE, yadL, yddB, yejO, or ynbB) were also screened again in isolation as described above. We cannot exclude the possibility that such transposon mutants have polar or second-site defects; however, the data are presented as a means of validating the negative selection detected by STM in the absence of competition between coscreened mutants.

All mutants were recovered in lower numbers (at least 0.5 log₁₀ CFU/g) than the parent strain tested in parallel in one or more of the organs sampled, except for the yddB::Tn and yejO::Tn mutants in the lung and the ΔcadA::kan^R, ΔfucP::kan^R, and ynbB::Tn mutants in the liver (Fig. 3). The absence of negative selection of yddB::Tn and yejO::Tn mutants in the lung is consistent with the phenotypes detected by STM (Table 1). The ΔevgS::kan^R, ΔyccC::kan^R, and ΔfliC::kan^R mutants were recovered in significantly lower numbers from the lung than the parent strain by 24 h postinoculation (Fig. 3A) (P ≤ 0.05). In the liver, significantly lower numbers of the entF::Tn, exoP::Tn, ΔfliC::kan^R, ΔhtrA::kan^R, traE::Tn, and ΔyccC::kan^R mutants than of the parent strain were obtained (Fig. 3B) (P ≤ 0.05). All mutants except for the ΔcadA::kan^R and yejO::Tn mutants were recovered from the spleen in significantly lower numbers than the parent strain (Fig. 3C) (P ≤ 0.05).

• Open in new tab
• Download powerpoint

Fig 3

Recovery of APEC strain χ7122 mutants from the lung (A), liver (B), and spleen (C) 24 h after intra-air sac inoculation of turkeys. The recovery of the ΔevgS::kan^R and ΔyccC::kan^R mutants relative to that of the parent strain was tested in 4 birds per strain. In a separate study, other mutants were tested relative to the parent strain in three birds each. Asterisks indicate a significant difference (P ≤ 0.05) between the concentration (log₁₀ CFU/g) of the mutant strain recovered and that of the parent strain.

Role of the group 4 capsule.A defined ΔyccC::kan^R mutant, predicted to lack the Etk tyrosine kinase required for the synthesis of the group 4 (O-antigen) capsule (64), proved to be highly attenuated. The group 4 capsule has been proposed to mask surface structures in E. coli O157:H7 and influences the colonization of the large intestines of rabbits (80), consistent with the negative selection of a signature-tagged ymcC mutant of E. coli O157:H7 in the intestines of calves (60). To the best of our knowledge, our data provide the first evidence that the group 4 capsule is required during systemic E. coli infection. The absence of the ymcDCBA-yccZ-etp-etk genes from the APEC O1 strain highlights the variable repertoire of virulence-associated genes among APEC isolates.

Role of flagella.While flagella have been implicated in the intestinal colonization of 10-day-old chickens and systemic virulence in day-old chicks after oral dosing with APEC O78 (28), our data with a defined ΔfliC::kan^R mutant confirm that flagella play a significant role upon respiratory challenge. Further studies are required to determine whether this reflects their activity as an adhesin per se, as proposed for H6 and H7 flagella of attaching and effacing E. coli (81–83), their role in bacterial motility, or their ability to act as agonists of innate immunity.

Role of the EvgAS system.Attenuation of the ΔevgS::kan^R mutant is consistent with the negative selection of three χ7122 evgS transposon mutants here and an APEC O2 evgA mutant by STM in an early lung colonization model in chickens (21). The evgAS genes encode a two-component sensory system that controls the transcription of numerous genes in E. coli, including some associated with acid resistance (84, 85). Such transcriptome studies were conducted in E. coli K-12, and it is feasible that APEC-specific genes may also fall under the control of this system.

Role of iron acquisition genes.The entF and yddB genes are predicted to be involved in iron uptake. Transposon mutants were negatively selected in the liver and spleen to a greater extent than in the lung in the primary screen and when tested in isolation, as may be expected given the need to acquire iron during systemic spread and the upregulation of genes involved in the synthesis and transport of siderophores in an APEC O1 strain exposed to chicken serum (72). EntF forms part of multienzyme complex (EntDEF) that mediates amide linkage of 2,3-dihydroxybenzoic acid l-serine during the synthesis of enterobactin. A recent study indicated that enterobactin secretion, but not enterobactin synthesis per se, is essential for the full virulence of APEC O78 strain χ7122 in chickens: a ΔentS mutant was attenuated in an intra-air sac coinfection model (33). A χ7122 ΔentD mutant tested in that study was negatively selected in lung and liver tissue relative to a virulent ΔlacI strain, but not significantly so (33). YddB is a predicted TonB-dependent porin for iron uptake, although formal evidence for such a role is lacking.

Role of Yad fimbriae.The yadL::Tn mutant was recovered at levels at least 0.5 log₁₀ CFU/g lower than those for the parent strain at each site sampled, although only the data for the spleen were significant. Taken together with the independent isolation of 5 mutants with defects in the yadN-ecpD-htrE-yadMLKC cluster by STM, the data suggest that Yad fimbriae play a subtle role in APEC pathogenesis. The yad cluster in APEC strain χ7122 is conserved in other sequenced APEC strains, E. coli K-12 MG1655, UPEC CFT703, and EHEC EDL933 and encodes fimbrial subunits and a chaperone-usher secretion system homologous to those involved in the production of E. coli type 1 fimbriae. Type I fimbriae are known to play an important role during the infection of chickens with strains of APEC serogroups O2 (27) and O78 (28). The predicted promoters of the E. coli K-12 yad operon and of the orthologous genes in E. coli O157:H7 (loc2; ECs0139-0145) have been reported to be inactive during growth in vitro (65, 86). However, forced expression of the yad operon in a Δfim mutant of E. coli K-12 led to the production of long, flexible pili, sometimes in bundles, and increased biofilm formation at 30°C but not adherence to epithelial cells (65). The yadN gene is enriched in human UPEC isolates relative to commensal E. coli strains, and Yad fimbriae are required for adherence to bladder epithelial cells and biofilm formation (87). They appear not to play a key role in bladder colonization in a murine urinary tract infection model but may act in concert with Ygi fimbriae in vivo (87).

Roles of other factors.Our data also support roles for the periplasmic serine protease HtrA (DegP), which mediates the response to stress and unfolded proteins, and exoP, which encodes a predicted prophage-encoded exonuclease family protein of ill-defined function. The significant attenuation of the traE::Tn mutant in the liver and spleen, but not in the lung, when tested in isolation is consistent with the phenotype of the mutant determined by STM screening. The data support a role for the pChi7122-1-encoded conjugation system, even though it is truncated relative to sequenced F plasmids (44). No significant attenuation could be detected for mutants lacking cadA (encoding a lysine decarboxylase that mediates stress responses in several enteric bacteria [reviewed in reference 88]) or yejO (a predicted autosecreted adhesin). Although both mutants were recovered in lower numbers than the parent strain at all three sites, mutants lacking FucP (l-fucose permease) or YnbB (a putative CDP-diglyceride synthase) were significantly attenuated only in the spleen.

Concluding remarks.Our analysis of the repertoire and function of APEC ST23 serogroup O78 genes in poultry reinforces the complexities of defining the APEC pathotype. Sequence typing and phylogenetic analysis of the core genomes of APEC indicate that E. coli strains causing the same disease of poultry arise from distinct lineages. Previous studies revealed that APEC ST95 O1 strains cluster with human ExPEC strains, prompting numerous investigations into the zoonotic potential of APEC. Our studies indicate that E. coli strains pathogenic for poultry are not uniformly related to ExPEC strains, since ST23 serogroup O78 strains clustered not with APEC O1 and UPEC strains but with a human enterotoxigenic E. coli strain of the same lineage. In turn, this challenges us to consider whether simple descriptors of E. coli strains as extraintestinal or diarrheagenic are supported by their phylogeny. We do not suggest that the similarity of the ST23 serogroup O78 strains studied is indicative of the potential for APEC strains from this lineage to cause human diarrheal disease, because they lack key enterotoxins. Rather, we infer that the core genome of the ST23 lineage can give rise to variants adapted to cause either avian or human disease via variation in the accessory genome. It is noteworthy that many of the attenuating mutations occurred in core chromosomal genes rather than in the variable plasmids, and saturating mutagenesis studies are needed to determine which features of “successful” backbones mediate E. coli fitness in vivo.

The evolutionary distance between APEC ST95 and ST23 is reinforced by the dramatic lack of conservation of virulence-associated genomic islands and the acquisition of distinct sets of virulence genes. This is further reflected in the lack of overlap between the virulence genes identified by STM in χ7122 and those identified in APEC ST95 by STM or targeted mutagenesis. Our analysis of signature-tagged and defined mutants supports roles for a number of factors found in the ST23 O78 strains (e.g., group 4 capsule, ETT2) but absent from the APEC ST95 O1 strain and suggests that some factors may exert their effects in a niche-specific manner.

This study represents the first high-resolution comparison of APEC genomes and provides valuable data for the rational design of strategies to control an important avian disease. It is clear from our work, and from preceding studies on the conservation of specific genes among field isolates, that extensive whole-genome analysis of strains from dominant APEC sequence types will be needed if we are to appreciate the full extent of diversity among these heterogenous pathogens. As the volume of sequence data for APEC expands, mutagenesis studies of the kind we report in target animals will be essential in order to determine which genes or combinations thereof are necessary for host adaptation and pathogenesis.