The Escherichia coli Hemolysin

RODNEY A. WELCH

[SECTION EDITOR: ALISON O’BRIEN]

November 23, 2005

Department of Medical Microbiology and Immunology, University of Wisconsin—Madison, Madison, Wisconsin 53706

Phone: 608-263-2700, Fax: 608-262-8418, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

AN OVERVIEW OF THE ESCHERICHIACOLI HEMOLYSIN

The E. coli hemolysin, earlier often referred to as the α-hemolysin, is the best-characterized repeats in toxin (RTX) toxin secreted by a type I exoprotein secretion system. The E. coli hemolysin is a significant virulence factor in murine models of peritonitis and ascending urinary tract infection (53, 78), which suggests it is likely to be an important cytotoxin in human, extraintestinal E. coli diseases. The classic laboratory phenotype for the E. coli hemolysin is the production of a zone of β-hemolysis (clearing) surrounding colonies on a sheep red blood agar plate. Structurally, the RTX toxins are dissimilar to the prototypical multimeric A-B subunit toxins. No apparent enzymatic activities are directly associated with the hemolysin protein. Hemolysin tertiary or quaternary structure(s) responsible for initiation or stabilization of its interaction with host epithelial or immune cell molecules are unknown. In addition, little information is available on direct analyses of its physical structures beyond determination of its primary sequence (23). The primary cytotoxic event appears to be the formation of a plasma membrane pore that leads to disruptive Ca²⁺ ion influxes (7, 49, 50).

The type I extracellular secretion system in gram-negative bacteria was originally called the RTX family because the members first described were the hemolysins and leukotoxins produced by E. coli, Pasteurella haemolytica, Actinobacillus species, and Bordetella pertussis (76). Following their discovery, other functional classes of exoproteins were found to share the classic nonapeptide Ca²⁺-binding repeats, C-terminal secretion signal, and dedicated three-gene transport system. The extracellular transport of RTX proteins requires an ABC-type inner membrane protein (B protein), a periplasm-spanning channel protein (D protein), and a TolC homolog (79). Besides the E. coli hemolysin and related toxins, the type I secretion family encompasses exoproteins with activities involving nodulation, proteolysis, lipolysis, heme-binding, or bacteriocin production (1, 19, 21).

Many of the RTX toxins have toxic activities limited to specific types of host cells and animal species (e.g., P. haemolytica leukotoxin). The E. coli hemolysin is broadly active toward many different cell types (erythrocytes, leukocytes, endothelial cells, etc.) (14, 34, 41, 69) from a wide variety of species, including humans and ruminants. Among E. coli or Salmonella strains there are no known examples of strict RTX leukotoxins in which lytic activity is limited to white blood cells. However, the pO157-encoded EhxA RTX toxin produced by E. coli O157:H7 strains has a mixture of cellular and host specificities unlike any of the other RTX toxins (5). Currently there is little direct evidence linking EhxA expression to the pathogenesis of E. coli O157:H7 hemorrhagic colitis and hemolytic-uremic syndrome.

With the discovery of a chromosomally encoded RTX toxin in epidemic strains of Vibrio cholerae (45) and Vibrio vulnificus (15) it has become apparent that the earlier consensus structure and genetic features of the toxin branch of the type 1 family require reevaluation. The general gene organization of the V. cholerae RTX locus is similar to that seen with either of the E. coli hly and ehx loci with C, B, and D RTX homologs, clearly indicating it is a member of the RTX family. However, the apparent A gene product lacks the canonical nonapeptide repeats. Instead it possesses an 18-residue glycine- and aspartate-rich repeat, which, although longer than the typical RTX repeat, can be modeled to form a Ca²⁺-binding β-roll motif that will be described in detail later in the review (6).

Early during the discovery of the RTX family it appeared that the RTX toxins were limited to extraintestinal pathogens (76). However, with the discovery of the E. coli O157:H7 EhxA and V. cholerae RtxA toxins it became evident that RTX toxins are likely to provide a selective advantage for pathogens infecting practically any niche in the body. This point continues to be relevant with the recent discovery of RTX-like gene sequences in the genomes of different E. coli pathotypes and Salmonella strains. For example, there are RTX B, D, and TolC-like genes linked to very large open reading frames (ORFs) encoding RTX-like proteins in Salmonella enterica serovar Typhi CT18 (ORF STY4458, accession no. AAL23085). These RTX-like proteins have lengthy tandem-arrayed repeats that are different from those seen for either the E. coli hemolysin or V. cholerae RtxA proteins. The functions of these candidate RTX exoproteins are unknown, but the absence of a linked RTX C-like gene suggests that these particular proteins are not likely to undergo a posttranslational fatty acid modification process. This suggests that they lack the cytotoxic activities typically attributed to the acylated RTX toxins.

E . COLI HEMOLYSIN STRUCTURAL DETAILS

Aside from the canonical RTX Ca²⁺-binding repeats, the 110-kDa, 1,023-amino-acid residue E. coli hemolysin polypeptide, HlyA shares many other common structural and chemical features with the broader toxin family. The E. coli hemolysin is abundant in positively charged amino acids in its amino-terminal third, whereas the distal third is rich with negatively charged amino acids. The significance of this general chemical property is unknown; however, in models of RTX interaction with host cells it is speculated that the N-terminal portion of RTX toxins forms a domain that initiates cell contact via an ionic interaction with the negatively charged plasma membrane (75). There is a long stretch of generally hydrophobic amino acid residues from position 200 through to 450. It was originally hypothesized that the hydrophobic region contained α-helices capable of forming a plasma transmembrane pore responsible for the primary cytotoxic mechanism (22, 54). An alternative hypothesis is that 10 amphipathic α-helices within the hemolysin, including six that occur in the 200- to 450-residue area, insert but do not traverse the lipid bilayer (62). This intramembrane insertion causes the displacement of phospholipids within one layer of the membrane leading to bilayer destabilization with pseudopore-like activity. Surprisingly, after much research by many different laboratories, the physical definition of the structural pore has yet to be achieved.

The extracellular targeting sequence recognized by the dedicated HlyB, HlyD, and TolC export machinery occurs within the C-terminal 60-amino-acid residues. This region has been the object of extensive studies (16, 24, 38, 65, 80). The hemolysin export sequence cannot be recognized as a specific primary sequence, but it is structurally defined only in that it possesses an amphipathic α-helix followed by a charge-rich linker sequence of 8 to 10 residues with the α-helix positioned approximately 45 residues from the C terminus (28).

As discussed above, posttranslational acylation of the RTX protoxin by HlyC-like proteins (51) is a common requirement for cytotoxic activities. For at least the E. coli hemolysin and the B. pertussis hemolysin/adenylate cyclase, the sites of these modifications are known (25, 29, 44, 58). Hackett et al. found that the B. pertussis hemolysin/adenylate cyclase RTX toxin is uniformly palmitoylated at CyaA_K983 (25). In the recombinant E. coli background, the position is palmitoylated 87% of the time, and the remaining molecules are myristoylated. In addition, in the E. coli background, a second position, CyaA_K860, is palmitoylated 60% of the time. The functional consequences of these acylation variations are that the CyaA prepared from the E. coli background is less erythrolytic but similar to the toxin purified from the Bordetella background in levels of invasive adenyl cyclase-mediated toxicity (26). In contrast, in in vitro reactions using proHlyA and HlyC, active HlyA can be produced that is modified with saturated C₁₄, C₁₆, and C₁₈ fatty acids. C₁₄ in the reaction provides the highest level of erythrolytic activity (29). Quite surprisingly, however, when native, in vivo modified HlyA prepared from either the Goebel or Welch laboratories is examined, the actual fatty acids used at positions 563 and 689 are saturated 14-, 15-, and 17-carbon amide-linked side chains (44) found on 68%, 26%, and 6% of the HlyA molecules, respectively. The modification of HlyA with odd-numbered carbon fatty acids is a unique observation. It also indicates that the in vivo HlyA species are very heterogeneous in structure. The functional significance of the different fatty acid species is unknown. The suspicion is that they may contribute to host and cell type specificity or interactions in host cell lipid-modified signaling pathways.

There are a few known requirements for hemolysin modification (25, 55, 64). Lysines are modified with fatty acids with a glycine residue just N-terminal to each lysine. A short α-helix is also predicted to be just N-terminal to the modified lysine (25, 55, 64). The biochemistry of the RTX toxin acylation event has been studied with E. coli HlyC and HlyA (64, 66, 71, 72). HlyC uses an acyl-acyl carrier protein (acyl-ACP) as the donor of the fatty acids. HlyC takes the fatty acids from the acyl-ACP donor to form either an acyl-HlyC intermediate (71) or noncovalent ternary acyl-ACP-HlyC-proHlyA complex (63).

The RTX toxin repeats are responsible for Ca²⁺ binding, which serves to explain the early observation that hemolysis or cytotoxicity by RTX toxins is Ca²⁺ dependent (11, 12, 27, 60). Nonetheless, the functional role of the Ca²⁺-repeat structure for RTX toxins remains obscure. Conformational changes occur when Ca²⁺ is added to HlyA prepared in Ca²⁺-free media or where HlyA is treated with EGTA (3, 27, 57). Mutational loss of a few of the repeats can be tolerated or compensated for by increased levels of Ca²⁺ ions (47, 58). The cellular association of RTX toxins requires the repeats and Ca²⁺, although hemolysin mutant proteins lacking most of the repeats can still weakly associate with susceptible cells (4). By analogy with the RTX alkaline protease from P. aeruginosa, the RTX toxin repeats are likely to form an elongated parallel β-roll structure with the β-strands wound in a right-handed spiral with Ca²⁺ ions bound at the turns between the strands (6).

It is critical to point out in summary that four acylation-independent events occur in E. coli hemolysin biogenesis and cytotoxicity. They are Ca²⁺ binding to the repeats, extracellular secretion, pore formation in artificial lipid bilayers, and loose association of the hemolysin with target cells (4, 7, 12, 46, 49, 51). Disappointingly, despite efforts by many different laboratories, the evidence is scant as to the secondary, tertiary, and multimeric structures for the E. coli RTX toxins. The absence of a resolved structure for HlyA has thwarted the structure and functional analyses of large sets of hemolysin mutants collected in many laboratories.

MECHANISM OF HEMOLYSIN CYTOTOXICITY

Despite numerous studies, the mechanism for hemolysin cytotoxicity remains unresolved. It is unlikely that erythrocytes are a significant target for the hemolysin during E. coli disease; however, many critical observations have been made with this convenient reagent. The significance of these observations for elucidation of important hemolysin structural and functional activities is uncertain. It is clear that the E. coli hemolysin causes rapid ion fluxes in treated cells. Jorgensen and colleagues (33) found that, within 2 min of treating erythrocytes with the E. coli hemolysin, Ca²⁺ ions begin to accumulate intracellularly and K⁺ ions are released from the cell into the medium. The ion flow appears to be selective because several different externally applied, small radiolabeled molecules (e.g., ³²PO₄) do not enter the cell. It is also certain that the final destruction of erythrocytes is by colloidal osmotic lysis (8). For other, more relevant, host cells the most popular hypothesis explaining hemolysin cytotoxicity is that a large sharp and unregulated Ca ²⁺influx initiates cytoskeletal destruction with multiple necrotic sequelae (reviewed in reference 75). Biochemical and physical definition of these events has not taken place. A recent observation by Uhlen and coworkers adds a new layer of complexity to the Ca²⁺ ion influx mechanism. They observed that semipurified hemolysin preparations taken from culture supernatants cause a continuous, low-frequency oscillation in intracellular [Ca²⁺] in primary rat renal tubular cells (73). Based on the use of specific inhibitors, the oscillatory effect is attributed to Ca²⁺ ion influx through L-type calcium channels and intracellular stores controlled by inositol triphosphate (13) and not a hemolysin-mediated transmembrane pore. The calcium oscillatory behavior induces proinflammatory cytokines, interleukin-6 (IL-6) and IL-8. The Ca²⁺ ion oscillatory effect occurs over a small, 10-fold concentration range of hemolysin where, below a threshold of input toxin, intracellular [Ca²⁺] does not change but at too high a hemolysin concentration the cultured cells lyse.

Although the hemolysin causes necrotic death at high concentrations that is characterized as lysis, apoptosis-like events can be observed at a lower, narrow window of "sublytic" concentrations for many different RTX toxins (35, 36, 39, 48, 67). If RTX toxins are apoptotic agents, it is not clear that they act by recognized apoptotic mechanisms. Classic tumor necrosis factor receptor (Fas)- or Bcl-2-mediated apoptotic cell death occurs 2 to 3 days after the induction event with intermediary signs of apoptosis, such as DNA fragmentation occurring nearly a day after induction. On the other hand when cells are treated with sublytic amounts of RTX toxins, apoptotic processes such as DNA fragmentation occur within 3 to 6 h (48, 67, 68). The difficulty lies in assessing the set of toxin concentrations and time conditions under which the RTX toxins are likely to act in vivo. It is possible that hemolysin-mediated cellular cytotoxicity is a mixed event in which, during the presence of an active infection and presumably continuous hemolysin production, there are instances and sites where sufficient amounts of hemolysin cause apoptosis and, in other instances, higher hemolysin concentrations lead to necrosis.

HEMOLYSIN CELLULAR RECEPTORS

Helena Ostolaza and colleagues have obtained results consistent with the hypothesis that the abundant erythrocyte surface molecule glycophorin acts as a receptor for the E. coli hemolysin (17). In addition, the same group has shown that the small, C-terminal 22-amino-acid hemolysin segment from residues 914 to 936 acts as the toxin ligand for glycophorin (18). Lally and coworkers demonstrated that the host cell surface leukocyte function-associated antigen (LFA-1) participates in RTX toxin activity (40). Anti LFA-1 antibodies were shown to neutralize cellular toxicity of the Actinobacillus actinomycetemcomitans leukotoxin. The role of LFA-1 in this toxin's action was further supported when a normally non-LFA-1-expressing and A. actinomycetemcomitans leukotoxin- or E. coli hemolysin-insensitive human erythroleukemia cell line, K562, was transformed with a recombinant plasmid encoding the two subunits (CD11a and CD18) of LFA-1, leading to toxin sensitivity. These results are consistent with LFA-1 acting as receptor for these two RTX toxins. The attraction of the RTX toxin ligand-LFA-1 receptor hypothesis is that sequence divergence among RTX toxins and the CD11a and CD18 subunits could account for the different host and cell type specificities observed for the RTX leukotoxins mentioned above. Jeyaseelan and coworkers performed a similar set of experiments with the P. haemolytica leukotoxin, which confirmed that bovine-specific forms of LFA-1 help mediate its toxicity (30). Although no new reports have supported LFA-1 in cytotoxicity of the E. coli hemolysin, our laboratory has observed that the SK-β₂-7 cell that fails to express LFA-1 because of a mutation in CD18 (74) is 32-fold reduced in sensitivity to the E. coli hemolysin (unpublished results). The reduced, but sustained, sensitivity of this cell line suggests that the E. coli hemolysin may use multiple, different receptor molecules or that possibly LFA-1 acts as a coreceptor where another molecule can bind significant amounts of the toxin. All in all, the results are consistent with the hypothesis that LFA-1 acts as a receptor for RTX toxins. In an alternative hypothesis that has not been tested, LFA-1 serves as a component of a cytoskeletal structure or as a component in a signaling pathway that RTX toxins disrupt.

HEMOLYSIN REGULATION

The regulation of the hlyCABD chromosomal operons commonly found among uropathogenic E. coli strains has been studied in detail, whereas little is known about specific elements of the regulation of the E. coli O157:H7 pO157-encoded ehxCABD genes. The hlyCABD genes are transcribed as a single transcript; however, at the end of hlyA, there is a substantial amount of transcriptional termination so that much more of the 5' hlyCA species exists than the full-length hlyCABD transcript (55). Apparently, this is the major mechanism responsible for adjusting the levels of the HlyB and HlyD transporters relative to their secretion substrate, HlyA (55). Despite strong conservation of nucleotide sequence for the hlyCABD genes from uropathogenic E. coli strains J96 and 536 genes studied, respectively, in the Welch and Goebel laboratories, the sequences immediately 5' to hlyC are not homologous and seem to be of different origins (37, 55). Although this observation suggested that there may be different transcriptional regulatory controls for the two versions of hlyCABD, major transcriptional control of the operons is mediated by a common transcriptional positive effector, rfaH. RfaH acts as an antiterminator at transcriptional pause sites early in the hlyCABD transcriptional unit (42, 43). The mechanism for rfaH control was recently described by Artsimovitch and Landick (2). An elegant set of experiments showed that RfaH binds to both RNA polymerase and the nontemplate strand of DNA at the pause site.

hhawas originally identified as a negative effector of the hlyCABD operon encoded by the large plasmid pHLY152, which was found in a murine fecal strain of E. coli (52). Recently, a leuX deletion mutant of uropathogenic strain 536 was shown to cause an increase in hha transcript levels and a subsequent reduction in hlyCABD transcript levels (20). This suggests that Hha activity may not be limited to the hly genes encoded on large plasmids as was originally proposed.

EPIDEMIOLOGY OF E. COLI RTX GENES

As described above, the two distinct versions of RTX toxin genes found in E. coli (hlyCABD and ehxCABD) have limited distribution among the different E. coli clonal lines or pathotypes. The well-studied E. coli hlyCABD genes are common to human extraintestinal isolates of E. coli that cause sepsis and pyelonephritis (32). The hemolysin occurs less frequently in cystitis strains and only rarely among normal fecal strains. Hemolytic E. coli strains are mostly limited to the ECOR B2 and D clonal designations (31).

Among the extraintestinal E. coli isolates, the hlyCABD genes were among the first virulence factors localized to unique, tRNA-associated segments of E. coli chromosomes. The hemolysin genes were eventually linked to P-type pilin and cytotoxic necrotizing factor-1 genes (9, 70). These large segments of virulence-associated genes were later coined pathogenicity islands (10). The chromosomal locations of the hlyCABD-containing islands vary among E. coli extraintestinal isolates (56, 77). In some animal and human enterotoxigenic E. coli isolates, hly genes closely resembling the chromosomal determinants can be found on large plasmids (61). The functional hemolysin genes are a clear example of significant horizontally disseminated genes mobilized and transferred among different clonal lineages of E. coli. The genetic mechanisms that mediate these transfers are unclear.

Among the human diarrheagenic types of E. coli, the ehxCABD version of the RTX toxin genes is limited to the common virulence plasmid, pO157. This plasmid is principally found among Shiga toxin-positive, O157:H7 isolates (59). The ehxCABD genes and the pO157 plasmid are not commonly found among other Shiga-like toxin-producing E. coli.

SUMMARY

For the past three decades the E. coli hemolysin has been the subject of research in many laboratories throughout the world. It has many fascinating genetic, biochemical, pathogenetic, and epidemiological features that are alluded to in this brief review. Recent progress with its study has admittedly slowed down because of the difficulty in deriving the physical structure of the hemolysin protein or other RTX toxins and establishing its precise cytotoxic mechanism and role in pathogenesis of extraintestinal E. coli disease. Genomic sequencing has revealed that there are additional RTX-like genes found among many different pathogens; perhaps new efforts to discover their functions will aid progress in the RTX toxin field.