In the United States during 2004–2005, dead redpoll finches from Alaska were submitted to the Alaska Department of Fish and Game. Three clinically healthy redpoll finches were trapped near the outbreak site. Standard necropsies included gross examination and collection of tissues into 10% neutral buffered formalin for histopathologic examination. Fresh tissues were frozen for microbiologic assays. Two other dead birds, a captive adult gyrfalcon (Falco rusticulos) from Idaho and a chicken (Gallus gallus) from Washington, were submitted for diagnosis to the Washington Animal Disease Diagnostic Laboratory in Pullman, Washington, USA.

In Canada in 2005, isolates were obtained from feces of clinically healthy redpolls and pine siskins (Carduelis pinus) trapped on Prince Edward Island. A total of 158 finches were sampled and included redpolls, pine siskins, and purple finches (Carpodacus purpureus).

In Australia in 2001–2002, isolates from birds were obtained from feces of clinically healthy domestic and trapped wild birds (8). Domestic birds included 9 chickens (G. gallus), 4 geese (Anser anser domesticus), 3 ducks (Anas platyrhynchos domesticus), and 1 guinea fowl (family Numididae). Wild birds totaled 634 birds representing 112 species.

In Scotland, isolates obtained during 1998–2000 from dead birds—Eurasian siskins (Carduelis spinus), greenfinches (Carduelis chloris), and chaffinches (Fringilla coelebs)—previously identified as E. coli O86:K61 (4), were obtained from M.J. Woodward (Veterinary Laboratories Agency Weybridge, Inverness, UK). Information about all isolates is summarized in Table 1.

To detect Enterobacteriaceae, we inoculated tissues, intestinal contents, or feces onto MacConkey agar and incubated the plates at 35°C. Isolated colonies were characterized by fermentation of lactose and glucose, production of oxidase, and production of indole from tryptophan. Additional biochemical characterization was performed by using a commercial kit (API 20E; bioMérieux, Hazelwood, MO, USA). E. coli serotyping was performed by the Gastroenteric Disease Center (Wiley Laboratory, Pennsylvania State University, University Park, PA, USA).

Genetic identification was based on 16S rRNA gene sequencing and/or PCR to detect housekeeping gene polymorphisms unique for the E. albertii/Shigella boydii lineage. 16S rRNA analysis was performed on 1 isolate from Alaska by sequencing >1,400 nt of the 16S rRNA gene (12). The amplicon was cloned into the pCR2.1 sequencing vector (TOPO TA Cloning Kit; Invitrogen, Carlsbad, CA, USA) and sequenced bidirectionally by automated dideoxy DNA methods. Partial 16S rRNA sequences of ≈500 bp of the 5′ end, including the V1, V2, and V3 variable regions (13), were determined for other isolates with the same primers, after which direct dideoxy sequencing was performed. A sequence similarity search was performed by searching the GenBank database with BLASTN.2.2.3 (14), and sequences were aligned with ClustalW2 (www.ebi.ac.uk/Tools/clustalw2/index.html). PCR was used to detect E. albertii lineage–specific genetic polymorphisms in the housekeeping genes lysP and mdh (10). As a positive control, PCR for the gene clpX, which is conserved in E. coli, Shigella, and the E. albertii/S. boydii lineage, was performed as described (10), except a corrected primer sequence for clpX_28 (5′-TGG CGT CGA GTT GGG CA-3′) (T.S. Whittam, unpub. data) was used. The negative control for the lysP and mdh PCR was E. coli strain DH10b.

PCR was used to test isolates for virulence genes found in Enterobacteriaceae—the central conserved region of intimin (eae, the attaching-and-effacing ligand), heat-stable enterotoxin (sta), and Shiga toxins (stx1 and stx2)—as described (15). Positive controls were E. coli strains S2 (for sta) and S14 (for stx1, stx2, and eae) from the Pennsylvania State University E. coli Reference Center. A multiplex PCR protocol that amplified the consensus portion of the B subunit of the cytolethal distending toxin gene (cdtB) as described by Toth (16) was modified by using each of the primer pairs (s1/as1 and s2/as2) individually to screen for cdtB in all isolates.

For sequencing eae, PCR primers that amplified ≈800 nt of the variable 3′ end of the eae gene were used as described (9). Because these primers did not work for all bird isolates, additional primer sequences were either taken from other studies or designed for this study (Tables 2 and 3). Sequence analysis was based on ≈726 nt in the 3′ variable region of eae, which corresponded to amino acids 33–275 within the C-terminal 280 aa of intimin (Int280). Nucleotide sequences were determined for each of the cdtB products obtained by using the s1/as1 (403 bp) and s2/as2 (411 bp) primer pairs (16). Predicted amino acid sequences for eae and cdtB were aligned with reference alleles by using the ClustalW method and MegAlign software (DNASTAR, Madison, WI, USA). Neighbor-joining dendrograms were constructed by using MEGA version 4 (18) with the p-distance metric and pairwise gap deletion.

MLST was performed on 26 isolates of E. albertii (Table 1) as described (10), with slight modification. Briefly, partial gene sequences for 6 conserved housekeeping loci (aspC, clpX, fadD, icdA, lysP, and mdh) were obtained by PCR and direct sequenced by automated dideoxy sequencing. Raw sequences were aligned by using Seqman Pro software (DNASTAR). Sequences for 11 E. albertii isolates from humans and 6 common E. coli pathotypes were obtained from www.shigatox.net. E. coli was used as an outgroup; strains used were enterohemorrhagic E. coli strain EDL933, Shigella flexneri strain 2747-71, enteroaggregative E. coli strain 042, enteropathogenic E. coli strain e2348/69, uropathogenic E. coli strain CFT073, and E. coli K-12. A neighbor-joining dendrogram was based on the concatenated nucleotide sequence and the maximum composite likelihood model by using MEGA version 4 (18). Details of the MLST procedure, including allelic typing and sequence type assignment methods, can be found at www.shigatox.net.

An overall phylogenetic representation of the genus Escherichia was generated by combining nucleotide sequence data from GenBank for Escherichia fergusonii with outgroup strains Salmonella bongori, S. enterica subsp. enterica serotype Typhi, and S. enterica subsp. enterica serotype Typhimurium. A neighbor-net network analysis was generated by using SplitsTree 4 software (19) (Figure 1, inset).

E. albertii isolates were compared by using a standard PFGE method (20) with minor modifications. Briefly, fragments of XbaI-digested bacterial DNA were separated in 1% agarose gel by using a CHEF-DR III PFGE apparatus (Bio-Rad, Hercules, CA, USA); pulse times were ramped from 2.2 to 54.2 seconds over 19 hours. Digital gel images were analyzed with Bionumerics software (Applied Maths, Sint-Martens-Latem, Belgium) by using the unweighted pair group method with arithmetic mean algorithm for cluster analysis of Dice similarity coefficients with a position tolerance of 2%.

Nucleotide sequences from this study were deposited in GenBank. Their accession numbers are EU926632–EU926649 and GQ140242–GQ140261.

Redpoll finches from the Alaska outbreak were typically found dead without obvious signs of disease. All those evaluated had adequate pectoral muscle mass, suggestive of acute death. Some had green fecal material pasted around their cloacae, suggestive of diarrhea. Gross lesions were inconsistent, but a few birds had darkened intestines distended with excessive yellow to green digesta. Histologic lesions were also inconsistent, but when present they were consistent with acute, severe, fibrinous, and necrotizing proventriculitis; multifocal heterophilic enteritis; and small-crypt abscessation. For some, gram-negative bacilli in large numbers were observed within the intestinal lumens. Attachment of bacteria to intestinal epithelial cells was not observed, although autolysis precluded assessment of the epithelium and ultrastructural studies to detect attaching-and-effacing lesions. No lesions consistent with septicemia were observed in any affected redpolls.

The affected gyrfalcon had appeared clinically healthy until found dead. Histologic examination failed to demonstrate enteric lesions, although evidence of septicemia was found. The chicken from Washington died after ≈1 week of illness, during which it appeared depressed and anorexic. Histopathologic examination showed severe, diffuse, necrotizing typhlitis; mild to moderate enterocolitis; and septicemia.

Bacterial cultures were performed for 8 dead redpolls from Alaska and 3 healthy redpolls trapped in the same area. Large numbers of non–lactose-fermenting gram-negative rods were isolated from the intestines and tissues of 5 of the dead redpolls but from none of the 3 healthy redpolls. Similar organisms were also isolated in large numbers from the tissues of the gyrfalcon and intestines of the chicken. In the healthy birds trapped on Prince Edward Island, non–lactose-fermenting bacteria were isolated from the feces of 11 (12%) of 95 siskins and 4 (12%) of 33 redpolls but from none of 30 samples from purple finches. From the healthy birds trapped in Australia, non–lactose-fermenting bacteria were isolated from 4 (18%) of 22 magpies (Gymnorhina tibicen), 1 (10%) of 10 honeyeaters (Melithreptus brevirostris), 1 (3%) of 38 wrens (Malurus cyaneus), 1 (7%) of 15 fantails (Rhipidura fulginosa), and 2 (22%) of 9 chickens.

The isolates were oxidase negative; fermented glucose but not lactose, sucrose, or xylose; produced indole from tryptophan; and were nonmotile at 35°C. Further biochemical characterization with the API 20E panel indicated that the isolates produced lysine decarboxylase and ornithine decarboxylase and fermented d-glucose, d-mannitol, and l-arabinose. The isolates did not utilize citrate; did not produce arginine decarboxylase, hydrogen sulfide, urease, tryptophan deaminase, acetoin, or gelatinase; and did not ferment inositol, l-rhamnose, d-sucrose, d-melibiose, or amygdalin. All isolates except that from the fantail from Australia (B1086) used β-galactosidase. Fermentation of d-sorbitol varied; it was not fermented by the isolates from finches from Alaska, Canada, and Scotland or the gyrfalcon or by 3 of the 9 isolates from birds in Australia (B1068, B1074, and B1086). API 20E testing identified the isolates that used β-galactosidase and fermented d-sorbitol as code 5144102 and weakly (43%) identified them as E. coli. Identical API 20E profiles were reported for the isolates identified as E. coli from the dead finches from Scotland (4). Although the isolates from Scotland were serotyped as O86:K61, an isolate from Alaska (1297-05-019) did not react with any of the 175 O E. coli antiserum samples, including O86. Variable use of β-galactosidase and fermentation of d-sorbitol resulted in API 20E codes of 5144502 or 4144102 and more robust identifications as E. coli (84% and 90%, respectively).

The nearly full-length 16S rRNA sequence of isolate 1297-05-19 from the redpoll in Alaska was most similar (1,470 [99.7%] of 1,475 nt) to sequences of E. albertii (AY696669) and S. boydii (1,467 [99.5%] of 1,475 nt, AY696670) isolated from humans. The 495-nt sequence of the 5′ end of 16S rRNA was determined for redpoll 5419-05-R from Canada, finch EC37098 from Scotland, and gyrfalcon 12055-07 and chicken 7991-07 from the United States. In all, 9 nucleotide polymorphisms were observed among these 5 isolates, including 5 clustered in the V1 region (98.2% overall identity). These sequences were 99.2%–99.6% identical to the sequence of an eae-positive strain of Hafnia alvei (Z83203), E. albertii from humans (AJ508775, AY696662–AY696664, AY696669), and S. boydii serotypes 7 and 13 (AY696670–AY696680). PCRs were positive for the E. albertii–specific alleles of lysP and mdh (10) in all isolates from birds. Collectively, these data tentatively identified these isolates as E. albertii.

All isolates from birds were positive for eae and cdtB but negative for stx1, stx2, and sta, the same repertoire of virulence genes reported for E. albertii isolates from humans (10). Alignment of the 3′ portion of the eae gene showed that the bird isolates possessed a variety of eae alleles, some novel and some similar to previously reported alleles (Figure 2, panel A). There was no clustering of bird eae alleles related to geographic origin, bird versus human origin, or isolation from diseased versus clinically healthy birds or humans. The largest cluster of bird alleles was found in representative isolates (Figure 2, panel A) from the redpolls from Alaska and Canada, the gyrfalcon and chicken from the United States, and the fantail from Australia, which were all nearly identical (1 nonsynonymous nt change each in the isolate from the redpoll and fantail from Alaska). This allele was distinct from other reference eae alleles and thus novel, but it was most similar to γ intimin. Alleles from the isolates from finch E37098 and siskin EC74699 from Scotland and magpie B101 from Australia were identical to each other but were also novel alleles most closely related to the μ allele in E. coli (17). The alleles from isolates from other wild birds and chickens from Australia were similar to previously reported allelic subtypes, including ε, α, and ν (17). Only the isolates from chickens B1068 and B1074 in Australia had an allelic subtype, ν 1.1, previously reported for an E. albertii isolate from a human (10,17).

All isolates tested (Figure 2, panel B) were PCR positive for cdtB with the s1/as1 primer pair. With the exception of the chicken from the United States and 5 isolates from birds in Australia (2 chickens [B1068 and B1074], 1 fantail, 1 honeyeater, and 1 wren), all were also positive for cdtB with the s2/as2 primer pair. Because these 2 primer pairs are specific for different types of cytolethal distending toxin (16), at least some isolates from birds appeared to carry multiple cdtB genes. Sequencing and alignment of the s1/as1 PCR products showed these to have ≈91% nt and 92% aa identity. On the basis of amino acid polymorphisms (Figure 2, panel B), avian cdtB alleles amplified by the s1/as1 primers were most similar to type II (birds from North America and Australia) and types III and V (the gyrfalcon, finches from Scotland, other birds from Australia) reference alleles (21–23). Sequencing and alignment of the s2/as2 PCR product showed that the sequences from the isolates from Australia were identical to each other, that the sequences from the isolates from North America and Scotland were identical to each other, and that these 2 groups of sequences were similar to each other with ≈99% identity at both the nucleotide and amino acid levels. These sequences were similar (1- or 2-aa differences) to the type I cdtB reference allele (24). The presence of a type I cdtB is consistent with the previous finding of a type I cdtB in the isolates from the finches from Scotland, identified by type-specific PCRs (16).

MLST of nucleotide variation at 6 loci (a total of 3,165 bp) in the genomes of isolates (Table 1) showed 3 main clades of E. albertii (EA 1, EA 2, and EA 3 in Figure 1). Isolates did not appear to cluster on the basis of host disease status (healthy, with diarrhea, or dead) or host type. All isolates from birds from North America were closely related and clustered in clade EA 2, along with 3 isolates from birds from Australia (honeyeater, wren, and fantail) and an isolate from a human with diarrhea (I2005002880 #36). The isolate from chicken 7991-07 was slightly divergent from the rest of the isolates from North America (5 synonymous and 1 nonsynonymous nt changes) and was indistinguishable from isolate I2005002880 #36 from the human. Isolates from the dead redpolls from Alaska, healthy finches from Canada, and the gyrfalcon were identical. The isolate from finch EC370-98 from Scotland was distantly related to other bird isolates and clustered with an isolate from a human with diarrhea (M2005000616 #8) in clade EA 1.

MLST strongly supported the biochemical and other molecular data indicating that the bird isolates in this study were E. albertii. Collectively, these isolates represent a distant relative of E. coli, a divergent lineage in the genus Escherichia, and novel diversity within the E. albertii species (Figure 1, inset).

PFGE showed that isolates from the 2 bird death epornithics (in Alaska and Scotland) each formed a clonal group (Figure 3), which suggests that these events were associated with expansion of a single clone from either a common source or bird-to-bird transmission. Overall, the PFGE banding patterns and dendrogram indicate that the isolates from birds and humans constitute a heterogeneous group, consistent with the heterogeneity identified in eae and cdtB and by MLST.