Characterization of the Roles of Hemolysin and Other Toxins in Enteropathy Caused by Alpha-Hemolytic Escherichia coli Linked to Human Diarrhea

Abstract

Escherichia coli strains producing alpha-hemolysin have been associated with diarrhea in several studies, but it has not been clearly demonstrated that these strains are enteropathogens or that alpha-hemolysin is an enteric virulence factor. Such strains are generally regarded as avirulent commensals. We examined a collection of diarrhea-associated hemolytic E. coli (DHEC) strains for virulence factors. No strain produced classic enterotoxins, but they all produced an alpha-hemolysin that was indistinguishable from that of uropathogenic E. coli strains. DHEC strains also produced other toxins including cytotoxic necrotizing factor 1 (CNF1) and novel toxins, including a cell-detaching cytotoxin and a toxin that causes HeLa cell elongation. DHEC strains were enteropathogenic in the RITARD (reversible intestinal tie adult rabbit diarrhea) model of diarrhea, causing characteristic enteropathies, including inflammation, necrosis, and colonic cell hyperplasia in both small and large intestines. Alpha-hemolysin appeared to be a major virulence factor in this model since it conferred virulence to nonpathogenic E. coli strains. Other virulence factors also appear to be contributing to virulence. These findings support the epidemiologic link to diarrhea and suggest that further research into the role of DHEC and alpha-hemolysin in enteric disease is warranted.

Escherichia coli is one of the major causes of human infectious diseases, partly because of the wide variety of virulence mechanisms and pathotypes (15), and new pathotypes continue to be described. A new pathotype was proposed by Gunzburg et al. after examining diarrheal pathogens in a prospective community-based study among Australian Aboriginal children (22). One group of isolates was significantly (P < 0.05) associated with diarrhea, and these isolates were particularly common among children younger than 18 months. The isolates did not produce any recognized enterotoxin or classic enteric virulence factor, although they exhibited diffuse or aggregative adhesion in a modified adhesion assay (15). All isolates were able to detach HEp-2 cell monolayers and were termed “cell-detaching E. coli.” This property was shown to be mediated by alpha-hemolysin, and we demonstrate below that all cell-detaching E. coli strains produce alpha-hemolysin and that some may also produce cytotoxic necrotizing factor 1 (CNF1) and other toxins. However, neither alpha-hemolysin nor CNF1 has been clearly demonstrated to be an enteric virulence factor, and the role of hemolysin in particular is controversial. We will refer to these isolates as diarrhea-associated hemolytic E. coli (DHEC) isolates.

Alpha-hemolytic E. coli strains have been associated with human enteric disease, especially among young children (8, 10–12, 20–22), and the related enterohemolysin of E. coli O157 (35) appears to be involved in enteric disease. There has, however, been no large prospective case-controlled epidemiologic study of the association of alpha-hemolysin with human diarrhea. Alpha-hemolytic bacteria are also associated with enteric disease and diarrhea in pigs, cattle, and dogs (9, 13, 33, 36, 44, 45). Porcine diarrheal strains are almost universally hemolytic (23a), and alpha-hemolysin in these isolates enhanced virulence and colonization (37) but was not itself diarrheagenic. More recent studies have found that Hly⁺ CNF1⁺ strains caused fluid accumulation in piglets (33) and that neonatal pigs were susceptible to challenge with Hly⁺ CNF⁺ strains, which caused bloody diarrhea, enterocolitis, and systemic disease (45).

In contrast, some earlier studies were unable to demonstrate a role for hemolysin in enteric disease, since neither hemolytic bacteria nor their supernatants caused fluid accumulation in ileal loops (10, 14, 37). Hemolytic strains may be isolated from the feces of asymptomatic people (26), and, among humans, hemolysin is more commonly associated with strains causing extraintestinal infections (5, 26).

The genetics and in vitro mechanisms of alpha-hemolysin are well known. The hlyCABD operon encodes the structural 110-kDa hemolysin protein (HlyA) and proteins involved in processing and export (42). Once secreted, hemolytic activity is short-lived, and this has complicated studies of hemolysin toxigenicity (42). Hemolysin does not require a receptor to bind to target cells, inserting instead into the target cell membrane to form a pore that allows the free flow of cations, sugars, and water. This leads to leakage of intracellular contents and affects the cytoskeleton and metabolism (4, 9, 42, 43). In extraintestinal infections, hemolysin has multiple effects and roles, including resistance to host defense, tissue damage, and lethality, either by direct action or by stimulation of inflammatory mediators and signal transduction pathways (7, 9, 16, 42).

CNF is a 114-kDa protein with homology to a family of dermonecrotic toxins (18) and is encoded by the monocistronic cnf gene, which lies just downstream of hly. The CNF1 toxin causes HeLa cells to become large and multinucleated as a result of actin disassembly, which results from activation of Rho (10, 19, 31). Similar to alpha-hemolysin, the role of CNF1 in diarrhea remains unclear. CNF1-producing strains have been isolated from diarrheal stools and have been associated with several outbreaks in humans (8, 10) and animals (13, 33, 44). Unfortunately, no large, prospective, case-controlled studies have been performed, and the best evidence for the pathogenicity of CNF1-toxigenic isolates is the marked virulence in piglet challenge experiments (45), outlined above. Purified CNF1 did not show enterotoxic potential in the suckling mouse or induce fluid accumulation in the rabbit ileal loop (10, 14), in contrast to the related CNF2, which is linked to enteric disease in animals (13, 14, 30). Both CNF toxins are extremely lethal, and have a variety of in vivo effects including tissue necrosis and edema (12–14).

In this paper, we characterize DHEC isolates that were obtained from a study where alpha-hemolysin was significantly associated with disease (22) and show that they are able to cause disease in rabbits. Using molecular genetics, we attempt to analyze the role of each gene in pathogenesis.

MATERIALS AND METHODS

Bacterial strains and plasmids.

A collection of 177 DHEC strains were obtained as part of a study of diarrheal virulence factors (22), and 3 of these isolates were selected for further study (Table 1). Also listed are mutant variants of wild-type DHEC and plasmids including cloned constructs of hly and cnf genes from strain A70.1 and strains used in mutagenesis.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant properties	Reference
Wild type
A70.1	Hly, CNF1, DA, MSHA, pap, detaching-cytotoxin positive, spindle factor positive	22
A98.1	Hly, CNF1, pap	22
55.3	Hly, CNF1	22
J96	Hly, CNF1, pap	29
Control
C600	Avirulent K-12	34
Mutant variants of 55.3
SE368	55.3 CNF1−	This paper
SE371	55.3 Hly−	This paper
SE372	55.3 Hly−, CNF1−	This paper
Plasmids
p3E1	hlyI cnf cosmid clone	This paper
p3E3	hlyI cnf cosmid clone	This paper
p4D3	hlyII cosmid clone	This paper
pSE376a	Hly+ CNF+ 28-kb SalI fragment from p3E1	This paper
pSE377a	Hly+ 15.3-kb partial PstI fragment from p3E1	This paper
pSE378a	CNF1+ 4.3-kb EcoRI fragment	This paper
pSE379	13-kb EcoRI fragment, containing cnf	This paper
pJP5603	oriR6K, lacZ Km	32
pJP5608	oriR6K, lacZ Tet	32
pSE297	1.4-kb PstI-BglII fragment from pSE379 in pJP5603	This paper
pSE298	1.4-kb PstI-BglII fragment from pSE379 in pJP5608	This paper
pSE345	0.55-kb EcoRI fragment from pSE377 in pJP5608	This paper
pSE346	0.55-kb EcoRI fragment from pSE377 in pJP5603	This paper
pSE266	0.95-kb Sau3AI fragment from pSE377 cloned into pUC18	This paper
pBGS18	Kmr pUC18 derivative	38
pHC79	Apr Tcrcos
Miscellaneous cloning strains
DH5α	supE44 ΔlacU169 (φ80 lacZΔM15) hsdR17 recA1 endA1 gyrA96 hi-1 relA1	34
JM109λpir	recA1 supE44 endA1 hsdR17 gyrA96 relA1 thi-1Δ(lac-proAB) F′ [traD36 proAB+ lacIqlacZΔM15] λpirR6K	32
S17-1λpir	thi pro hsdR hsdM+ recA::(RP4-2-Tc::Mu-Km::Tn7, Tpr Smr)λpir	32

^aThese plasmids were transformed into E. coli C600 for RITARD model studies. 

Assays for cell detachment, alpha-hemolysin, and CNF.

Cell detachment was assayed as described by Gunzburg et al. (22). A semiconfluent monolayer of HEp-2 cells in a 24-well tissue culture tray was washed three times with phosphate-buffered saline (PBS) supplemented with CaCl₂ and MgCl₂ (each at 0.01%) (PBS-CM), and 1 ml of PBS-CM was added to each well. A 10-μl volume of log-phase bacterial culture was added, and the plate was incubated at 37°C for 90 min under 5% CO₂. Monolayers were washed three times with PBS-CM, fixed with 70% methanol for 10 min, stained with 0.13% crystal violet for 10 min, and then briefly destained in water. Significant destruction of the monolayer was recorded as cell detachment and was quantified by eluting crystal violet with a solution of 50% ethanol, 49% water, and 1% sodium dodecyl sulfate (SDS) and measuring the absorbance of the eluate at 590 nm. Alpha-hemolysin was detected by the presence of characteristic zones of lysis on Columbia agar (Oxoid) containing 5% washed sheep erythrocytes, with observation at 4 h and again after overnight incubation, as described by Beutin (5).

CNF was detected by the method of Oswald et al. (30) by assaying the ability of freeze-thawed bacterial lysates to cause characteristic multinucleation, cytotoxicity, and morphological changes to HeLa cells. CNF activity was quantified as the 50% multinucleation titer (MN₅₀), the maximal dilution able to cause multinucleation in 50% of HeLa cells. Antibody neutralization of CNF activity was demonstrated by preincubation of the lysate overnight at 37°C with neutralizing antiserum before addition to HeLa cells. Antisera to CNF1 and CNF2 were kindly provided by A. Caprioli, Instituto Superiore di Sanità, Rome, Italy.

Molecular genetic techniques.

DNA preparation, cloning, manipulation, and Southern hybridizations were performed as described by Sambrook et al. (34) unless otherwise stated. Transposon mutagenesis with Tn1725 was performed by the method of Ubben and Schmitt (41). Chromosomal DNA was prepared by the method of Ausubel et al. (2).

To clone virulence factors from A70.1, chromosomal DNA of A70.1 was partially digested with Sau3AI and ligated into the BamHI site of pHC79. Concatamers were packaged into phage heads and transfected into E. coli HB101 as specified by the manufacturer (Amersham λDNA in vitro packaging kit). A total of 412 colonies were screened for cell detachment and alpha-hemolysin.

A 4.3-kb fragment containing cnf was sequenced by the method of Bankier et al. (3), and a library of contiguous, randomly sheared fragments was constructed. Double-stranded DNA sequencing was performed with the Advanced Biosystems Inc. (ABI) Prism dye terminator sequencing kit, as specified by the manufacturer, and an ABI automatic sequencer. The sequence was complied and analyzed with software packages MacVector for the Macintosh and tfasta, fasta, and align written for Unix.

Chromosomal mutagenesis by marker exchange.

Mutagenesis of chromosomal genes was performed by the method of Penfold and Pemberton (32). A DNA fragment internal to the gene of interest was cloned into pJP5603 or pJP5608, which contain a lacZ with a multiple-cloning site from pUC18 and a pir-dependent R6K origin of replication. Transformants containing the insert of interest were first screened in JM109λpir to enable blue-white selection and then transformed into S17-1λpir for conjugation into a rifampin-resistant (Rif^r) variant of the DHEC strain. Mutations were selected by loss of the corresponding phenotype and confirmed by Southern blotting. All Rif^r variants of DHEC strains used as recipients for conjugations were otherwise indistinguishable from the parent.

Animal challenge.

The RITARD (reversible intestinal tie adult rabbit diarrhea) assay was performed by the method of Albert et al. (1). Rabbits were challenged with 10¹⁰ bacteria grown on colonization factor antigen agar and were observed for up to 7 days for diarrhea and other symptoms and for shedding of challenge organisms. Shedding was monitored by performing daily rectal swabbing. Animals that developed frank diarrhea were sacrificed, and all the remaining animals were sacrificed on day eight. Following sacrifice, the rabbits were examined for gross pathological changes, and sections were taken from the midjejunum, the proximal and distal ileum, the proximal and distal colon, the cecum (excluding the blind segment that was surgically manipulated), the appendix, the rectum, and the mesenteric lymph nodes. These were preserved in buffered formal saline, sectioned, stained with hematoxylin-eosin, and examined by light microscopy (see above). Pathological changes including inflammation, polymorphonuclear leukocyte (PMN) infiltration, and peritonitis were noted semiquantitatively, ranging from normal (0) or mild (1) to severe (3).

Protein analysis.

Standard protein analysis, SDS-polyacrylamide gel electrophoresis, immunization, Western blotting, and immunostaining techniques were performed as outlined by Harlow and Lane (24), unless otherwise stated. Monoclonal antibodies to alpha-hemolysin (h2A, h11A, f11F, I1C) (25) were kindly provided by F. Hugo, Institute of Medical Microbiology, University of Giessen, Giessen, Germany.

For production of polyclonal antiserum to A70.1 hemolysin, alpha-hemolysin was precipitated from broth by the method of Bhakdi et al. (6) and subjected to SDS-polyacrylamide gel electrophoresis. The band of interest was excised from the acrylamide and macerated, and a 50% suspension in saline was injected intraperitoneally into male Swiss mice on a schedule of 1, 14, 21, and 28 days; the mice were bled on day 35.

RESULTS

Cloning and characterization of alpha-hemolysin from DHEC.

A cosmid library of A70.1 DNA was screened for production of alpha-hemolysin, and three clones were isolated. Restriction analysis (Fig. 1) revealed that two clones, p3E1 and p3E3, were identical and differed from a third clone, p4D3. From p3E1, hemolytic activity was subcloned on a 28-kb SalI fragment (giving pSE376 [Table 1]) and a 15.3-kb partial PstI fragment (generating pSE377). pSE377 was further mapped with probes, restriction enzymes, and transposon Tn1725 (41), which also confirmed the region containing hly (Fig. 2). Gene order was determined with the use of a series of DNA probes directed to different regions of the operon.

FIG. 1.

Map of the two chromosomal loci from A70.1 containing the hlyCABD operon as predicted from the maps of cosmids p3E1 and p4D3. The two copies of hly are designated hlyI (in p3E1) and hlyII (in p4D3). See the text for further details. B, BglII; Bm, BamHI; E, EcoRI; S, SalI; , hly; ░⃞, cnf.

FIG. 2.

Map of fragments in pSE377, which contains hly subcloned from cosmid p3E1 from A70.1 and is shown here aligned with previously mapped hly loci as obtained from reference 29 (a and b) and GenBank accession no. M10133 and M12863 (a). B, BglII; Bm, BamHI; E, EcoRI; P, PstI.

The hly operon of DHEC was compared to those previously described. The restriction map and gene order were very similar to those of the chromosomally located hly of uropathogenic E. coli (UPEC) J96 and the plasmid-borne hly from plasmid pHly512, isolated from an animal pathogen (Fig. 2). Four monoclonal antibodies raised against HlyA from UPEC recognized a 110-kDa supernatant protein produced by A70.1 and hly clones. A polyclonal antiserum raised against A70.1 HlyA recognized a similar 110-kDa band from A70.1, hly clones, and UPEC J96 (results not shown).

The remaining collection of 176 DHEC strains was examined by colony blotting, and they all recognized hly DNA probes (see below) and the polyclonal antiserum. Most, but not all, recognized at least one monoclonal antibody, indicating antigenic variation in HlyA among DHEC strains.

Cloning and analysis of CNF1 toxin production.

DHEC strains were examined for production of CNFs and other toxins on HeLa cells. Of 177 DHEC, 54 (30.5%) (Table 2) were found to produce CNF1 toxin based on characteristic morphological alterations (Fig. 3a). There was no evidence for production of CNF2 or verotoxins. In strains A70.1, A98.1, and 55.3, the identity of the toxin was confirmed by neutralization with specific antisera against CNF1, and in strain A70.1, it was confirmed by sequence analysis (see below).

TABLE 2.

Association of CNF1 toxin and probe among DHEC strains^a

CNF1 toxin	No. of strains with cnf probe		Total no. of strains
	Present	Absent
Present	54	0	54
Absent	6	117	123
Total	60	117	177

^aDHEC strains were examined for production of CNF1 toxin as outlined in Materials and Methods. Reactivity to the Sau3AI probe for CNF1 (from plasmid pSE266) was determined by colony hybridization. Specificity, 90%; sensitivity, 100%; positive predictive value, 90%; negative predictive value, 100%. 

FIG. 3.

Toxin production by CDEC A70.1. (a) A70.1 bacterial cell lysate, diluted 1:20 and inoculated onto HeLa cells. In addition to generalized cell toxicity, two types of abnormal cells are observed: (i) large, multinucleated cells with diffusely staining borders, characteristic of CNF1; and (ii) densely staining, mononuclear, highly elongated cells, referred to as spindle cells. Magnification, ×1,000. (b) Dilution (1/80) of A70.1 lysate. Multinucleated cells are few, but more spindle cells are observed, and they predominate in some fields. Magnification, ×1,000. (c) Dilution (1/20) of lysate from DH5α(pSE260), containing the cloned cnf gene from A70.1 on a high-copy number plasmid. All cells are multinucleated. Spindle cells are absent. Magnification, ×1,000.

CNF1 activity was encoded on cosmid clones p3E1 and p3E3 but not p4D3 and was subcloned from p3E1 as a 28-kb SalI fragment (pSE376) and a 13-kb EcoRI fragment (pSE379). Figure 4 shows the map of pSE379, illustrating the positions of cnf and a Tn1725 insert (described below). The restriction map of the cnf region was similar to that in UPEC EB35 (17). Using Tn1725, which contains an EcoRI site in the terminal repeat regions, we were able to insert an EcoRI site that enabled us to isolate cnf on a 4.3-kb fragment. This subclone, pSE378, was highly active in toxin assays, with an MN₅₀ of 1:1,280 compared to 1:40 to 1:80 for A70.1 (Fig. 3c) and strains 55.3 and A98.1. This subclone was sequenced as a series of overlapping random fragments. Analysis of the 4,295-bp sequence (GenBank accession no. A42629) and the predicted translation product indicated significant DNA and amino acid homology (99.6% amino acid identity) to cnf from UPEC (accession no. X7670) (18). Differences occurred in the 5′ region of cnf and upstream of the gene.

FIG. 4.

Map of pSE379 containing a 13-kb gene fragment including cnf, showing the position of the Tn1725 insertion that added the EcoRI site. Also shown are subclones derived from pSE379. pSE378 is a 4.3-kb EcoRI fragment cloned into pUC18. pSE266 is a 0.95-kb Sau3AI fragment cloned into pUC18 and used as probe for cnf. B, BglII; E, EcoRI; P, PstI; S, SalI; Sm, SmaI.

A probe for cnf was made by cloning a 902-bp Sau3AI fragment (corresponding to nucleotides 1061 to 1963) from pSE378. This probe hybridized to 60 of the 177 DHEC strains, including all the strains producing assayable toxin (Table 2), and is 90% specific and 100% sensitive for predicting CNF1 toxin production.

Identification of two chromosomal hly loci in A70.1.

The restriction map of cosmids p3E1, p3E3, and p4D3 was determined with the assistance of hly and cnf probes. This demonstrated that the identical cosmids p3E1 and 3E3 carried cnf while p4D3 lacked cnf and possessed different BglII and BamHI sites. Identically sized fragments were observed in Southern blots of chromosomal DNA, which confirmed that A70.1 contained one cnf gene and two hly operons. These were named hlyI, which is present on p3E1 and linked to cnf, and hlyII, which is present on p4D3. Probing of chromosomal DNA from strains 55.3 and A98.1, as well as mutagenesis, suggested that these two isolates also possess two hly genes and one cnf gene.

Mutagenesis of chromosomal genes in wild-type DHEC.

To understand the role of hly and cnf in intestinal diseases, chromosomal genes were inactivated by marker exchange techniques. The CNF1 gene was inactivated by cloning a 1.4-kb PstI-BglII fragment (corresponding to nucleotides 1044 to 2495 of the cnf open reading frame) into pJP5603, an oriR6K-based suicide vector (Fig. 5). After this recombinant plasmid (pSE298) was conjugated into A70.1 Rif^r, 80% (38 of 48) of the transconjugants had lost the ability to produce CNF1 toxin. Insertion into the chromosome was demonstrated by disruption of the 23-kb BamHI fragment. The same technique on Rif^r variants of A98.1 and 55.3 inactivated CNF1 production in 100% of transconjugants.

FIG. 5.

Derivation of fragments for recombinant suicide vectors. A 1.4-kb BglII-PstI fragment was cloned from cnf into pJP5603, giving pSE297, or into pJP5608, giving pSE298. A 0.55-kb EcoRI fragment was cloned from within hlyA into pJP5603 or pJP5608, yielding pSE346 and pSE345, respectively. B, BglII; Bm, BamHI; E, EcoRI; S, SalI.

To inactivate hly by this strategy, we proposed to use suicide vectors with different antibiotic resistance markers to inactivate the two hly genes. A 500-bp EcoRI fragment from within hlyA from hlyI was cloned into pJP5603, generating pSE346, which inserted into and inactivated cloned hly genes on plasmid pSE377 (from hlyI). When pSE346 was introduced into A70.1, it inserted into the chromosome at hlyII in 20/20 attempts. When other fragments were cloned from hlyI into vectors pJP5603 or pJP5608, the resultant recombinant suicide vectors always (30 of 30) inserted into hlyII in A70.1, as seen in Southern blots of chromosomal DNA. To inactivate hlyI, a tetracycline-resistant analog of pSE346 was constructed by cloning the 500-bp EcoRI hlyA fragment from hlyI into pJP5608 (a Tc^r derivative of pJP5603). In A70.1, this plasmid (pSE345) inserted into the chromosome at hlyII and was never observed to insert into hlyI. When pSE345 was introduced into A70.1 derivatives containing the insertion of pSE346 into hlyII (i.e., A70.1 hlyI hlyII::pSE346), all 1,400 Tc^r Km^r transconjugants screened were hemolytic, indicating a failure to insert into and inactivate hlyI. These data collectively indicate that hlyI in A70.1 is resistant to mutagenesis by this strategy, possibly due to the tertiary structure of the DNA.

This strategy was then attempted on A98.1 and 55.3 without success. The lone exception was isolated after pSE346 was introduced into 55.3, generating a nonhemolytic variant, SE371 (Table 1). SE371 produced CNF1 and was otherwise indistinguishable from the parent, and Southern blotting demonstrated that hlyI had been insertionally inactivated but that the pathogenicity island containing hlyII had spontaneously excised. Given the difficulty in constructing a nonhemolytic variant by genetic techniques, this spontaneous mutant was used in further manipulations and was used to construct a double mutant defective in both alpha-hemolysin and CNF1. The CNF1-targeting pSE297 was introduced into SE371, and CNF1 production was lost in 1 (2%) of 42 strains examined. This strain, SE372 (Table 1), lacks both Hly and CNF1 production. This and the other variants were used in animal challenge experiments.

Other toxins produced by DHEC.

A number of potentially novel toxins appear to be produced by some DHEC strains, as observed in the effect of freeze-thawed bacterial lysates on HeLa cells in the 72-h CNF assay. It was demonstrated that the following phenotypes were not due to alpha-hemolysin, since they were not mediated by the cloned hly from A70.1 or observed in most other DHEC strains.

The first toxin caused detachment of HeLa cells in the CNF1 assay. While CNF1 can cause limited death and detachment, 1:10- and 1:20-diluted lysates from several CNF1⁺ strains, including A70.1, exhibited a pronounced cell-detaching activity that was not observed in other CNF1⁺ strains. This phenomenon was not due to CNF1, since 55.3, A98.1, and A70.1 exhibited a similar MN₅₀ of approximately 1:40, implying similar levels of CNF1 production, yet only A70.1 caused a complete loss of the HeLa cell monolayer. Further, lysates from DH5α containing the cloned cnf of A70.1(pSE378) exhibited potent multinucleating activity (MN₅₀, 1:1,250) but did not cause detachment. Finally, mutation of cnf in A70.1 abolished multinucleation but not detaching activity. In addition, two DHEC isolates that did not produce CNF1 (as determined by multinucleation of HeLa cells and DNA probe) were cytotoxic. These results demonstrate that cell-detaching cytotoxicity exists separately from CNF1.

The second potentially novel toxin caused unusual morphological alterations to HeLa cells. This phenotype was first identified in lysates from A70.1, which caused (in addition to the cell enlargement, loss of border definition, and multinucleation characteristic of CNF1) some HeLa cells to become thin and elongated, a phenomenon referred to as spindle cells (Fig. 3). Dilution of A70.1 lysates led to reduced multinucleating and enlarging activity but had less effect on spindle-forming activity, and at a 1:80 dilution, spindle cells dominated some fields (Fig. 3b). This activity was not observed in the lysates from the cnf clone (Fig. 3c). This demonstrates that CNF1 activity is distinct from spindle-forming activity. However, insertional inactivation of cnf in A70.1 caused a marked reduction in the spindle-forming activity of lysates. Although it is possible that CNF1 potentiates the spindle-forming factor, it appears unusual that two toxins with completely opposite toxic effects could act synergistically. CNF1 leads to large, rounded cells with diffusely staining cytoplasm, while spindle cells are extremely elongated and mononuclear, with a normally staining cytoplasm.

Animal challenge experiments.

Animal models were used to evaluate DHEC enteropathogenicity. Oral inoculation of streptomycin-treated specific-pathogen-free mice and rabbit ileal loops were attempted. While some colonization and pathological changes were observed, virulence was difficult to reproduce and was often mild or moderate, and so these models were abandoned (results not shown).

In the RITARD model, DHEC strains exhibited virulence with a set of clear, unique pathologic findings. Rabbits infected with DHEC exhibited depression, cramping, and diarrhea which ranged from mild to frank (Table 3). All rabbits that developed frank diarrhea produced mucoid diarrhea with anal staining. Some rabbits died. At sacrifice, there was evidence of fluid accumulation in the small and large intestines, sometimes with mucus and blood. Infected animals showed a number of marked histological changes in both the small and large intestines (Fig. 6). These included multifocal areas of mucosal erosion and necrosis, hyperemia, and colonic goblet cell hyperplasia. The level of hyperplasia was remarkable, and the mucosa in rabbits infected with strain A70.1 was observed to be 2 to 3 times the normal thickness. Other inflammatory changes included activation of the Peyer’s patches, PMN infiltration into the lamina propria and lumen, edema, and lymphocytic hyperplasia. However, the pathologic findings varied with each rabbit and infecting organism. In contrast, rabbits infected with the negative control E. coli C600 did not develop diarrhea and their intestines were entirely normal. It is clear that DHEC strains are more virulent in RITARD than is C600 and, overall, that DHEC strains are significantly (χ², P < 0.05) associated with diarrhea. To compare histopathological changes, outcomes were scored semiquantitatively (Table 4). By using this approach, it was evident that there was considerable variation between DHEC strains but that, overall, DHEC strains were significantly (t test, P < 0.05) associated with inflammation and that more inflammatory changes were observed in the large intestine.

TABLE 3.

Development of diarrhea, death, and shedding of organisms in a RITARD model challenge with DHEC

Challenge strain	Rabbit no.	Diarrhea			Shedding (day stopped)	Comments
		Day started	Day stopped	Maximum severity
Wild typec
A70.1	1	—a	—	—	—	Dead on day 2
	2	1	1	Mild	4
	3	2	2*b	Frank	2*
	4	—	—	—	6
	5	1	2*	Frank	2*
A98.1	6	—	—	—	—	Dead on day 1
	7	—	—	—	6
	8	1	2*	Frank	2*
	9	1	1	Mild	2
	10	1	1	Mild	5
55.3	11	—	—	—	—
	12	—	—	—	—
	13	3	3*	Frank	3*
	14	3	3*	Frank	3*
Mutant variants
SE368 (55.3 CNF−)	15	—	—	—	3	Dead on day 5
	16	1	1*	Frank	1*
	17	—	—	—	4	Dead on day 6
	18	1	1*	Frank	1*
SE371 (55.3 Hly−)	19	1	1*	Frank	1*
	20	1	3*	Frank	3*	Moderate on day 1, frank on day 3
	21	—	—	—	4
	22	1	1*	Frank	1*
SE372 (55.3 Hly− CNF−)	23	—	—	—	7
	24	2	2*	Frank	2*
	25	3	3*	Frank	3*
	26	3	4*	Frank	3*	Slight on day 3, frank on day 4
	27	1	2*	Frank	2*	Slight on day 1, frank on day 2
Recombinant plasmids in C600
pSE376 (C600 Hly+ CNF+)c	28	—	—	—	2
	29	—	—	—	—
	30	1	3*	Frank	3*	Slight on day 1, frank on day 3
	31	3	3*	Frank	3*
pSE377 (C600 Hly+)c	32	3	3*	Frank	3*
	33	4	5*	Frank	5*	Slight on day 4, frank on day 5
	34	6	7*	Frank	7*	Slight on day 6, frank on day 7
	35	4	6*	Frank	6*	Slight on day 4, frank on day 6
pSE378 (C600 CNF+)	36	—	—	—	2
	37	1	2*	Slight	2*	Died in lab accident
	38	—	—	—	—
	39	2	2*	Frank	2*
	40	—	—	—	4
Negative control C600
	41	—	—	—	—
	42	—	—	—	—
	43	—	—	—	—
	44	—	—	—	—

^a—, no disease was observed. 

^bAsterisk indicates that rabbit was sacrificed on day indicated. 

^cSignificant associations with diarrhea (χ² test, P ≤ 0.05); CDEC wild type versus C600, Hly⁺ clones in C600 versus C600, C600(pSE377) versus C600. 

FIG. 6.

Histopathological changes to rabbit intestines after challenge with A70.1. (a) Multifocal areas of mucosal erosion with submucosal accumulation of neutrophils and lymphocytes, capillary congestion, and lacteal dilatation. Rabbit ileum stained with hemotoxylin and eosin. Magnification, ×400. (b) Diffuse presence of active germinal centers in Peyer’s patches along the length of the ileum with a marked increase in the number of intraepithelial lymphocytes (arrow) and tingible-body macrophages. Rabbit ileum, stained with hemotoxylin and eosin. Magnification, ×400.

TABLE 4.

Summary of findings in RITARD model challenges with DHEC^a

Strain	Virulence factor(s)	No. of rabbits	No. dead	No. with diarrhea	% with diarrhea	Diarrhoeal severity (mean)b	Mean duration of diarrhea (days)c	Inflammation in small intestined	Inflammation in colond	Ratio of inflammation (large:small intestine)
A70.1	CNF+ Hly+	5	1	3	75	62.5	6.8	38	150	4
A98.1	CNF+ Hly+	5	1	3	75	50	4	262	187	0.7
55.3	CNF+ Hly+	4	0	2	50	50	4	32.4	38	1.2
SE368	55.3 CNF−	4	2*	2	50	50	7	75	125	1.7
SE371	55.3 Hly−	4	0	3	75	75	10	125	50	0.4
SE372	55.3 CNF− Hly−	5	0	4	80	80	8.8	20	20	1
pSE376	CNF+ Hly+	4	0	2	50	50	5.5	25	50	2
pSE377	Hly+	4	0	4	100	100	6.5	100	125	1.25
pSE379	CNF+	5	0*	2	40	30	3.8	0	40	NDe
C600	CNF− Hly−	5	0	0	0	0	0	4	7	1.75

^aDead rabbits excluded, except when marked with an asterisk. In those cases, two rabbits infected with 55.3 CNF⁻ died late in the experiment but had not developed diarrhea earlier. These rabbits were regarded for calculations as two nondiarrheal rabbits and were not excluded from calculations as for other deaths. One rabbit infected with pSE260 died from noninfectious causes after developing mild diarrhea. This rabbit was included and was assumed to have suffered mild diarrhea for 7 days, for purposes of calculation. 

^bThe maximal severity of diarrhea in each rabbit, where mild = 50 and frank = 100, was added, and the total score was divided by the number of rabbits. A score of 100 indicates that all rabbits developed frank diarrhea. 

^cThe number of days (in 7 days) of diarrhea × the severity of diarrhea (mild = 1, frank = 2) for each rabbit was added and divided by the number of rabbits. 

^dInflammation was recorded if inflammation, PMN infiltration, or peritonitis was observed. Histopathological descriptions of inflammatory reactions were graded 0, 1, 2, and 3, and the total score was divided by the number observed and adjusted for the number of rabbits observed. The small intestine score consists of scores from the jejunum, proximal and distal ileum; while the large intestine included proximal and distal colon. 

^eND, not done. 

To examine the role of different virulence factors in DHEC-mediated enteric disease, we compared rabbits infected with either wild-type DHEC strains, isogenic mutants, C600 containing cloned DHEC virulence genes, or C600. All rabbits infected with a C600 strain containing hlyI cloned on a high-copy-number vector developed frank diarrhea and showed marked inflammatory reactions in both small and large intestine. This clone was significantly more (χ², P < 0.05) likely to cause diarrhea than was the plasmid-free variant, suggesting that cloned hly is able to confer virulence upon nonpathogenic E. coli strains. Rabbits infected with the CNF1⁺ clone developed diarrhea yet were not statistically more likely to develop diarrhea than were controls (χ², P < 0.05). They also exhibited intestinal inflammation, although less than that observed in rabbits infected with the hemolytic clone.

Paradoxically, the loss of either hly or cnf from DHEC 55.3 did not lead to a measurable reduction in disease, as measured by diarrheagenicity or histopathological changes, and these strains remained more virulent than C600. However, loss of both CNF1 and hemolysin production was associated with low intestinal inflammation (Tables 3 and 4). These results suggest that other virulence factors are present in DHEC strains and complicate our understanding of the role of hly in DHEC-associated diarrhea.

DISCUSSION

DHEC strains were initially described as a class of E. coli that were significantly more common in children with diarrhea than in controls (22). To examine the diarrheagenic ability of DHEC in vivo, we used the RITARD model, in which the effect of each strain on the entire gut of a rabbit is examined for up to a week. In this model, DHEC strains caused frank mucoid diarrhea and a set of clear, unique, and marked intestinal pathologic findings in both the small and large intestines, including necrosis, hyperplasia, and multiple indicators of inflammation. In contrast, avirulent E. coli C600 caused no pathologic effects or diarrhea in rabbits.

The mechanism by which DHEC strains caused enteric disease was unknown as they lacked traditional diarrheal virulence factors (22). All DHEC strains produced alpha-hemolysin, and we have demonstrated that DHEC hemolysin could not be significantly distinguished from UPEC hemolysin on the basis of the restriction map, DNA hybridization, protein size, or antibody reactivity. Since the role of hemolysin in enteric disease is unresolved, we sought other toxins among DHEC strains that may explain their apparent diarrheagenicity. Approximately one-third of isolates produced CNF1 toxin, as demonstrated by the toxin assay and DNA hybridization. The analysis of cnf in A70.1 indicates that it is very similar to cnf from UPEC. There is evidence in a few DHEC isolates for two novel toxins which are phenotypically and genotypically distinct from alpha-hemolysin and CNF1. The first toxin is a HeLa cell cytotoxin found in A70.1 and some other DHEC strains. The second toxin observed in A70.1 possessed spindle-forming activity, since bacterial lysates caused a subset of HeLa cells to become densely staining, thin and elongated. While this was not mediated by cnf, mutation of cnf in the wild type markedly reduced spindle-forming activity. It is possible that CNF1 is necessary to potentiate toxin activity or, more likely, that mutation has had polar effects on downstream genes. It is not known what is encoded downstream of cnf.

Since the majority of DHEC strains do not appear to produce toxins other than hemolysin and CNF1, we examined the role of these factors in DHEC-induced disease with wild-type strains, clones, and mutant variants. The evidence from cloned hly supports its role as a virulence factor. When hly was cloned on a high-copy-number vector, it conferred on avirulent C600 the ability to cause diarrhea in rabbits and other pathologic changes. Notably, hemolysin appeared to be associated with inflammation, especially in the colon. By using statistical tests, the presence of both plasmid pSE377 (hly) and the hly gene was significantly associated with diarrhea. In contrast, the cloned cnf gene could not be demonstrated statistically to cause diarrhea, and so its role, if any, in this model of disease appears to be minor.

Complicating this analysis, however, is the data obtained from isogenic mutants of DHEC 55.3. Mutagenesis of cnf, hly, or both did not significantly affect the onset, duration, or severity of diarrhea, although loss of both was associated with a decrease in intestinal inflammation. This indicates that factors in strain 55.3 other than cnf or hly may also mediate diarrhea in the RITARD model, suggesting multiple virulence mechanisms.

In summary, the data suggests that DHEC strains are virulent and that alpha-hemolysin, the factor shared among all DHEC strains, is a virulence factor. This supports our initial epidemiologic observation linking hemolysin to diarrhea and agrees with other studies that found statistically significant associations and/or linkage to a particular outbreak (8–13, 20–22, 33, 36, 43, 44). Further, it supports the findings of Wray et al. (45), who demonstrated that Hly⁺ CNF1⁺ strains from pigs were virulent in piglets and caused pathologic changes generally similar to those observed by us, including diarrhea, death, and effects on both small and large intestines such as edema, inflammation, and necrosis. Rather than following classic secretory diarrhea, the disease was closer to that due to proinflammatory and invasive pathogens. The strains studied by Wray et al. (45) appeared to be more virulent in the piglet model, possibly reflecting their challenge of piglets with pig pathogens compared to our challenging of rabbits with human pathogens.

We propose several modes by which alpha-hemolysin could act as a diarrheal toxin. The first involves pore formation in the enterocytes, allowing the free flow of ions into the lumen. This may be enhanced by a Ca²⁺ flux into the cell, stimulating the arachidonic acid pathway and upregulation of secretion, or by generalized cytotoxicity, perturbing both secretion and absorbtion. Other pore-forming hemolysins shown to cause diarrhea include Vibrio cholerae El Tor hemolysin (39), Vibrio parahaemolyticus TDH (28), Serpulina hyodysenteriae hemolysin (40), and delta-hemolysin of Staphylococcus aureus (27). Because DHEC diarrhea appears to be associated with marked inflammation, this suggests that the inflammatory effects of hemolysin seen in extraintestinal infections and in vitro (7, 9, 16, 42, 43) are present in intestinal infection and may be important in diarrhea.

If we are to consider alpha-hemolysin to be an enteric virulence factor and DHEC strains to be diarrheal pathogens, it is possible that UPEC strains are also enterovirulent. Certainly, the virulence factors observed in DHEC strains did not distinguish them from uropathogens, and it is unclear if DHEC strains are uropathogenic organisms functioning as enteric pathogens or are specialized enteric pathogens. The common dogma that divides intestinal from extraintestinal E. coli pathogens may be an oversimplification that needs to be reexamined. However, it is possible that specific factors such as a specialized HlyA type are present in DHEC strains and are functionally different from those in UPEC strains.

It has been demonstrated that hly from different isolates may be more than 95% homologous but can differ markedly in the regulation of expression, HlyA stability, and virulence (4, 23). This was most elegantly shown in an animal model of UPEC pathogenesis (23). Deletion of a chromosomally encoded hly from a human UPEC led to a significant reduction in toxicity for mice, and reintroduction of hly cloned from the chromosomes of various UPEC strains on a recombinant plasmid restored toxicity. However, hly isolated from a plasmid of an animal enteropathogen led to a very marginal increase in toxicity despite being very highly related to hly of UPEC strains. Therefore, the marked similarities of DHEC hly to both UPEC hly and animal enteropathogen hly may nonetheless mask real and significant differences that are relevant to pathogenesis in this model. Indeed, we have observed with monoclonal antibodies that DHEC produce different HlyA subtypes that cluster with other virulence factors (unpublished data). This linkage suggests that there are DHEC subtypes that may cause different types of disease, with only certain categories responsible for diarrhea, and may explain why the epidemiologic link to diarrhea is often not as strong as that for “classic” enteropathogens.

Finally, we must address the findings of earlier workers, who were unable to demonstrate a role for hemolysin in enteric disease. As outlined above, the type of hly used may be a crucial factor. Thus, while DHEC hly may be a virulence factor, this is unlikely to apply to hly from all E. coli strains. Second, most studies have attempted to demonstrate diarrheagenicity in short-lived models appropriate for classic secretory toxins (e.g., cholera toxin) such as the rabbit ileal loop and have used toxin preparations despite HlyA being very labile. We would not expect that the type of disease observed with DHEC would yield a positive result in these studies, and we were unable to show virulence in the rabbit ileal loop. Marked disease was observed only several days after whole bacteria were inoculated into the RITARD model, and most pathologic findings were observed in the large intestine.

In conclusion, we believe that DHEC strains are potential enteropathogens and that the variant of alpha-hemolysin encoded on the chromosome is an important but by no means the only virulence factor in the RITARD model of infection. We believe that these preliminary results call for further experiments, including experiments with larger groups of animals and those with more precisely defined measures of pathogenesis. Finally, we believe that this result may eventually cause us to redefine the distinction between uropathogens and enteropathogens and how a factor that promotes virulence in one site may function in another. Certainly, the way in which UPEC interacts with the intestine may determine its ability to subsequently cause urinary tract infection.