INTRODUCTION

Top Previous Next References

Protein toxins are prominent virulence factors of many pathogenic bacteria. While toxins of gram-positive bacteria do not generally require activation, many toxins of gram-negative bacteria are translated in an inactive form and require a processing step, most often a proteolytic cleavage, to generate the active form. Enzymatic toxins such as Shiga toxin, cholera toxin, pertussis toxin, diphtheria toxin, and Pseudomonas aeruginosa exotoxin A undergo proteolytic cleavage to produce a catalytic A fragment acting in the eukaryotic target cell (279). Similarly, many nonenzymatic toxins that insert into eukaryotic membranes require proteolytic cleavage to allow oligomerization and pore formation; e.g., Vibrio cholerae El Tor hemolysin is cleaved at its N terminus, while Aeromonas aerolysin, Clostridium septicum alpha-toxin, and P. aeruginosa cytotoxin are cleaved at their C termini (13, 219, 236). The pore-forming hemolysin (HlyA) of Escherichia coli represents a unique class of bacterial toxins that require a posttranslational modification for activity, specifically the covalent amide linkage of fatty acids to internal lysine residues; the suggestion that activation of the Serratia and Proteus hemolysins ShlA and HpmA involves covalent modification has not been supported by subsequent experiments (123). After introducing the hemolysin toxin, this review will focus on the posttranslational mechanism of HlyA modification (maturation) and will examine its relationship to protein modifications involving lipid groups that occur in prokaryotes, lower eukaryotes, and mammalian systems. The functional significance of protein lipidation in these examples will then be reviewed and used as a basis for discussion of the possible role(s) of acylation in the cytokine-inducing and pore-forming functions of the HlyA toxin.

HlyA is an important virulence factor in E. coli extraintestinal infections such as those of the upper urinary tract (81, 107, 124, 318). It is one of a close family of membrane-targeted toxins assumed or proven to be influential not only in urinary tract infections but also hemorrhagic intestinal disease, juvenile periodontitis, pneumonia, whooping cough, and wound infections; these toxins include enterohemorrhagic O157 E. coli hemolysin (257), the leukotoxin of Pasteurella haemolytica (291), the hemolysins and leukotoxins of Actinobacillus spp. (46, 52, 89, 173), the bifunctional adenylate cyclase-hemolysin of Bordetella pertussis (100), and the hemolysins of Proteus vulgaris (167, 319), Morganella morganii (167), and Moraxella bovis (101) (Table 1). These toxins have 30 to 75% sequence identity to E. coli HlyA and share (i) posttranslational maturation, (ii) a C-terminal calcium-binding domain of acidic glycine-rich nonapeptide repeats that has led to the RTX (repeat toxin) family nomenclature, and (iii) export out of the cell by type I secretion systems (34, 59, 129, 171, 320). The posttranslational modification is unique to this toxin family, but Ca²⁺ binding and type I secretion are both common to other bacterial proteins (18, 34, 86, 221, 301). In HlyA there are between 11 and 17 glycine-rich repeats. When Ca²⁺ is bound (one calcium ion per repeat), these form short -strands organized in an unusual "spring-like" structure called a parallel -barrel or -superhelix (18). Calcium binding is an absolute requirement for cytotoxic activity (36, 37, 187, 232) and occurs outside bacteria following export, the intracellular level of free calcium in E. coli being very tightly regulated to 0.1 µM (92), a level too low for HlyA activity.

The synthesis, maturation, and secretion of E. coli HlyA are determined by the hlyCABD operon (81, 132, 171, 225) (Fig. 1). The membrane-located export proteins are synthesized at a lower level than the cytosolic HlyC and pro-HlyA, in part due to transcription termination within the hlyCABD operon (81). This termination is suppressed by the elongation protein RfaH and a short 5' DNA sequence, ops (operon polarity suppressor) (7, 8, 60, 225, 311), that act together to allow the transcription of long operons such as hly, rfa, and tra that encode the synthesis and export of extracellular components important to the virulence and fertility of gram-negative bacteria (7, 9).

View larger version (22K): [in this window] [in a new window]   FIG. 1.   Hemolysin synthesis, maturation, and export by E. coli. (CM, cytoplasmic membrane; OM, outer membrane; P, total PMF). Expression of the hemolysin operon is governed by components upstream of the hly genes. The hlyA gene encodes inactive prohemolysin, which is activated by HlyC. HlyA is secreted by a type I process. Ca2+ binds to the glycine-rich repeats of the toxin externally before interacting with the mammalian membrane.

The pro-HlyA protoxin is matured in the cytosol to the active form by HlyC-directed fatty acylation (see below). The maturation increases the hydrophobicity of the protein but is not required for export (186). E. coli HlyA and its toxin relatives are all secreted across both membranes by the type I export process employing an uncleaved C-terminal recognition signal (223, 281) but no N-terminal leader peptide (82) or periplasmic intermediate (83, 168). The HlyA secretory apparatus comprises HlyB (an inner membrane traffic ATPase), HlyD (an inner membrane protein that is suggested but not shown to bridge to the outer membrane), and TolC (an outer membrane protein) (260, 313, 315). In E. coli and most other pathogens, TolC is encoded by a gene separated from hlyCABD, but in B. pertussis the toxin locus includes tolC (cyaE). Type I secretion signal sequences have been located within the C-terminal 24 to 80 amino acids (137, 168, 293). The HlyA C terminus is predicted to contain an amphipathic helix (159, 281), and circular dichroism and nuclear magnetic resonance spectroscopy of the HlyA signal has shown that -helices are formed in a membrane mimetic environment (330). Despite a lack of identity between their primary sequences, the interchangeability of the export genes suggests that higher-order structures in the signals are shared among the extended family of hemolysins, leukotoxins, and proteases (256, 262, 301, 329).

HlyB has an integral membrane domain fused to an ATP-binding cassette (traffic ATPase) cytoplasmic domain, which undergoes conformational change on ATP binding (170) and couples ATP hydrolysis to HlyA export (166). Topological models proposed from fusion data suggest that HlyB inserts in the cytoplasmic membrane via six helical transmembrane segments with both its N and C termini in the cytoplasm. The predicted HlyB cytoplasmic loops are large and positively charged, while the periplasmic loops are small (97). The function of HlyD is less clear, but it is proposed to have a single transmembrane segment with the N terminus in the cytoplasm and a large C-terminal region in the periplasm (260). TolC is a minor outer membrane protein believed to form ion-permeable channels. Electron microscopy of two-dimensional lattices of TolC in phospholipid bilayers has revealed it to be a trimeric porin-like structure with a C-terminal extramembrane domain believed to form a periplasmic bridge to the energized inner membrane components of the translocation complex (172). HlyBD-dependent type I secretion shares with the Sec secretion across the cytoplasmic membrane an early requirement for the total proton motive force (PMF) but also has a late stage that does not require PMF, membrane potential, or the proton gradient. A translocation intermediate identified in the PMF-independent late stage is closely associated with the inner membrane, possibly in a translocation complex spanning both membranes (169). The belief that HlyA interacts with HlyB is compatible with the finding that suppressor mutations in HlyB partially compensate for mutations in the HlyA secretion signal (273), and a view of the assembly of a transport-competent complex occurring in an ordered manner is now being substantiated by cross-linking experiments (34, 300). A consequence for HlyA secretion of the virtual absence of Ca²⁺ ions in the bacterial cytoplasm is that the glycine-repeat structures should be flexible, allowing translocation of the polypeptide across the bacterial membranes in an unfolded state.

BIOLOGICAL EFFECTS OF HLYA AND RELATED TOXINS

Top Previous Next References

Host Cell Specificities of the Toxin Family

The toxins of the HlyA (RTX) family have been grouped on the basis of their lytic and toxic effects on mammalian host cells: hemolysins exhibit little target cell specificity, while leukotoxins have pronounced species- or cell-specific effects (59, 320) (Tables 2 and 3). E. coli HlyA has a wide spectrum of cytocidal activity, attacking erythrocytes, granulocytes (28, 91), monocytes (29), endothelial cells (294), and renal epithelial cells of mice, ruminants, and primates (155, 212). The 61% identical hemolysin (Ehx) from enterohemorrhagic E. coli O157 lyses sheep and human erythrocytes and kills bovine leukocytes (albeit with a specific activity more than 20-fold lower than that of HlyA), but, unlike HlyA, it has virtually no activity against human leukocytes (16). These contrast with the P. haemolytica and Actinobacillus actinomycetemcomitans leukotoxins which have little erythrolytic activity but a potent cytotoxic activity toward phagocytic cells of particular animals: the P. haemolytica leukotoxin (LktA) lyses leukocytes only from cattle, sheep, and other ruminants (274), while the A. actinomycetemcomitans leukotoxin (AaltA) specifically lyses human and primate polymorphonuclear lymphocytes (328). A. pleuropneumoniae produces hemolysins (ApxIA and ApxIIA) and a leukotoxin (ApxIIIA) (89) (Table 2).

How such reported host cell specificities arose during the evolution of the toxin family is not clear from phylogenetic trees derived from DNA sequences (320), and the physical basis for cell specificity is unresolved. Studies performed on hybrid toxins created by the exchange of putative domains between HlyA, LktA, ApxIIA, and AaltA have indicated that specificity is unlikely to lie in a single discrete feature. Activity against erythrocytes has been correlated with a central region of HlyA and an N-terminal region of ApxIIA (88). The ability to kill leukocytes has been mapped to C-terminal regions of LktA and ApxIIA but appears less sharply defined in HlyA (88, 202). The critical region of AaltA required for recognition of human target cells spans its glycine-rich repeats (175), whereas the specificity of LktA for ruminants has been linked to a feature at its N terminus (88).

Ambiguity in the experimental definition of the target cell range may contribute to the lack of understanding of targeting by these toxins. Targeting has been defined by the cytotoxic end result (death), but this is the result of a multistage mechanism that might involve receptor recognition, membrane insertion, and protein-protein interaction. There are therefore many reasons why a particular cell line might be insensitive to a particular toxin. The toxin may not interact with the membrane, and this may occur for AaltA with human erythrocytes (298) and for LktA with nonbovine erythrocytes (15, 42). The toxin may adsorb to the membrane but fail to insert into the lipid bilayer, and this may be the behavior of inactive, nonacylated protoxins. Even if the toxin does permeabilize the membrane, the cell line will appear insensitive if it possesses a lesion repair mechanism. It is also possible that the bacterial host will affect the final toxin activity. The B. pertussis CyaA hemolysin synthesized in E. coli has a fourfold-lower hemolytic activity than does the native CyaA (109), and while ApxIIA protein secreted from its native host A. pleuropneumoniae is both hemolytic and cytotoxic, expression of the apxIICA genes in E. coli generates a toxin that is cytotoxic but has little or no hemolytic activity (305). LktA from E. coli carrying the P. haemolytica lktCA genes caused weak lysis of erythrocytes despite reports that LktA has a target cell range limited to leukocytes and platelets (87, 125). Continued biochemical investigation of the action of these toxins should establish whether continued division into hemolysins and leukotoxins is appropriate.

An explanation for the apparent cell specificity of these toxins might be offered by their recognition of specific membrane receptors. The existence of receptors for many bacterial toxins, both enzymatic and pore forming, is well established. Diphtheria toxin enters mammalian cells by using the epidermal growth factor-like growth factor as a receptor (220), and the AB₅-type toxins such as cholera toxin, Shiga toxin, and verotoxin bind to cell surface ganglioside lipids (279). The binding of Staphylococcus aureus pore-forming leukotoxins to human polymorphonuclear neutrophils is initiated via a calcium channel or a receptor linked to a calcium channel (280). Aerolysin of Aeromonas hydrophila binds to erythrocytes and T lymphocytes via high-affinity receptors that have been shown to be glycoproteins with common glycosylphosphatidylinositol (GPI) anchors (73); Clostridium perfringens enterotoxin also uses several structurally related proteins as receptors (152); and the receptor for C. difficile toxin A has been identified as sucrase-isomaltase (245). Whether there is a receptor for the E. coli hemolysin is uncertain. Dose-response binding assays performed by one group support an upper limit of 4,000 HlyA binding sites per erythrocyte, implying at least some degree of specificity (15). However, this contrasts with results from others, who estimated up to 50,000 toxin molecules bound per erythrocyte and, using radiolabelled hemolytically active HlyA, subsequently found no evidence for saturation in binding to erythrocytes and leukocytes, indicating that this process is not conditional on specific receptors (31, 76). The absence of a specific receptor could be the reason why HlyA can attack a wide spectrum of target cells; indeed, HlyA can even bind to synthetic planar lipid membranes, suggesting that binding is relatively nonspecific (203). Nonspecific adsorption to the cell surface, rather than high-affinity binding to a surface receptor, has also been proposed to precede membrane insertion of B. pertussis CyaA (133). Inactive (unacylated) precursor toxin does not inhibit cell binding by active, mature toxin for both HlyA and CyaA, compatible with there not being a saturable receptor (15, 133).

It is possible that receptor-ligand interactions do occur with some members of the toxin family, for instance LktA and AaltA, displaying a narrower range of target cell specificities. Competition between inactive LktA mutants and mature LktA has been demonstrated, suggesting that a receptor in the leukocyte membrane is involved in LktA target specificity (61), and LktA is unable to bind to bovine leukocytes pretreated with protease, suggesting the involvement of a proteinaceous component in toxin binding (42). AaltA has been suggested to use a -integrin as a receptor, since it is able to bind both subunits of the lymphocyte function-associated antigen type 1. HlyA may use the same ₂-integrin for recognition of human leukocytes, but no direct binding has been shown (176). Integrins, a large family of heterodimeric transmembrane receptors that are differentially expressed by a variety of cell types, have the potential variability to distinguish both cell type and species (except in erythrocytes) and are used as receptors during Shigella and Yersinia invasion of mammalian cells (84, 85, 130). Protein sequences potentially involved in putative receptor recognition would appear most likely to lie within the nonconserved domains of the toxin molecules (do the differences between E. coli HlyA and Ehx define regions of interaction with a human leukocyte receptor?). A model could be put forward in which toxin action against erythrocytes is an intrinsic property and does not use a receptor while action against leukocytes is defined by specific receptor interactions. The involvement of target cell components such as integrins would be consistent with the lack of receptor involvement in erythrocytes.

Pathogenic bacteria frequently subvert host cell functions such as signal transduction pathways, cytoskeleton rearrangement, and vacuolar trafficking (reviewed in references 84 and 85). For example, enteropathogenic E. coli (EPEC) secretes proteins such as EspA and EspB that stimulate host phospholipase C, inducing inositol triphosphate production, the mobilization of intracellular Ca²⁺ stores, tyrosine phosphorylation of host proteins, and cytoskeleton rearrangement (251). Cell entry by Yersinia, Listeria, Salmonella, and Shigella also involves the activation of host signalling and a more extensive cytoskeleton rearrangement, which is controlled by small GTP-binding proteins belonging to the Ras family, namely, Rac, Rho, and CDC42. Shigella, Listeria, and Yersinia phosphorylate host proteins, including substrates for the nonreceptor tyrosine kinase Src. Salmonella, Mycobacterium, Chlamydia, and Legionella reside in intracellular membrane-bound vacuoles, and the stability of these vacuoles is believed to result at least in part from the bacteria interacting with vesicular trafficking that is normally mediated by a family of small GTP-binding proteins called Rabs. Proteins of the Ras, Src, and Rab families will be highlighted again in this review because of their possession of fatty acyl groups. There are also many examples of bacterial molecules inducing cytokine synthesis in eukaryotic cells, and although the most widely studied are lipopolysaccharide (LPS), capsular polysaccharides, and peptidoglycan, they also include porins, fimbrial proteins, heat shock proteins, and lipoproteins, as well as extracellular proteins such as proteases and toxins (reviewed in reference 120). Both gram-positive and gram-negative pathogens produce toxins that induce cytokine release; e.g., interleukin-1 (IL-1) and tumor necrosis factor (TNF) are induced in murine macrophages and human monocytes in response to listeriolysin O, pneumolysin, C. difficile toxin B, and Streptococcus pyogenes erythrogenic toxin A. Cholera toxin increases IL-6 synthesis and decreases TNF- production by rat peritoneal mast cells, pertussis toxin enhances IL-4 production, and verotoxin induces the release of IL-1, IL-6, and TNF- in mouse macrophages.

Erythrocytes exposed to E. coli HlyA rapidly undergo cytoskeleton rearrangement, resulting in the formation of teardrop-shaped projections from the surface (147). HlyA also affects cytokine production. At very low, sublytic concentrations, HlyA is a potent trigger of G-protein-dependent generation of inositol triphosphate and diacylglycerol in granulocytes and endothelial cells, stimulating the respiratory burst and the secretion of vesicular constituents (30, 103). HlyA stimulates the release of IL-1 and TNF from human monocytes, the lipoxygenase products leukotriene B₄ and 5-hydroxyeicosatetraenoic acid and nitric oxide from endothelial cells, and IL-1 (but not TNF-) from cultured monocytes, while it inhibits the release of IL-1, IL-6, and TNF- from human leukocytes (29, 102, 164, 295). When injected into mice, HlyA produces a cytokine response similar to those seen in vitro, elevating the levels of IL-1 and TNF in serum (200). This ability to affect cytokine release is shared by all the members of the HlyA toxin family; in vitro, LktA and AaltA reduce the lymphocyte response to various stimuli (195, 197), at low concentrations, LktA stimulates the production of IL-1 and TNF by bovine mononuclear phagocytes (287, 326) and potentiates the production of histamine and certain eicosanoids in response to chemokines (1). LktA and ApxII activate bovine and porcine neutrophils, respectively (63, 121, 306). These effects may occur without necrotic cell death, since HlyA and AaltA cause apoptosis of human lymphocytes and LktA causes apoptosis of bovine mononuclear cells and granulocytes (142, 197). As discussed below, some inflammatory cytokines possess a comparable acyl modification to that of HlyA.

E. coli HlyA and other RTX toxins alter the membrane permeability of host cells, causing lysis and death. Lysis of erythrocytes might provide bacteria with iron, and killing nucleated cells may prevent phagocytosis. The benefit to extracellular bacteria as a result of the induction of a host response is not so obvious. A consequence of the release of cytokines may be to induce inflammation, disrupt epithelial cell junctions, and favor translocation of bacteria through the intestinal barrier. The activation of host cell signal transduction pathways may also be advantageous to bacteria if the host responds by presenting a receptor used by them for binding. Such mechanisms are used by B. pertussis, where binding of filamentous hemagglutinin to a monocyte integrin causes the presentation of a second filamentous hemagglutinin-binding site (131), and by EPEC, where adherence to epithelial cells occurs after tyrosine phosphorylation by the host of an inserted bacterial protein, which then associates directly with intimin, an EPEC adhesin (160).

It is possible that some of the biological effects so far assigned to the HlyA toxin actually reflect cooperative responses to both toxin and LPS. In vivo and in vitro studies indicate that exposure to LPS, or mediators stimulated by LPS, increases the subsequent response to HlyA and LktA (261, 287). One response of host cells attacked by HlyA is the rapid and massive shedding of LPS receptors (CD14), which may then bind to neighboring cells and make them sensitive to LPS. Such a mechanism might underlie the long-range detrimental effects of pore-forming toxins in host organisms (312). The induction of a host response may not be an alternative to cell death but an alternative pathway to death, proceeding via programmed cell death or apoptosis instead of lysis. Many bacterial toxins as well as members of the HlyA family can induce apoptosis (126). Host cell death by apoptosis would occur without leakage of cellular components, without inflammation, and without damage to surrounding cells (4).

E. coli HlyA appears to have a two-stage interaction with eukaryotic membranes: a reversible adsorption sensitive to electrostatic forces and an irreversible insertion (10, 233). Transition to the inserted form is associated with a change of conformation (211) and is favored by membranes with a fluid state and low cholesterol content, which may explain earlier findings that membranes made from asolectin, a crude lipid mixture from soybean, are very sensitive to HlyA, in contrast to membranes composed of pure lipids such as phosphatidylcholine or phosphatidylserine (21, 231). Once inserted into a membrane, HlyA behaves as an integral protein; it cannot be extracted without the use of a detergent (27). HlyA causes target cell lysis by forming pores which display cation selectivity and voltage and pH dependence, as shown in experiments with whole cells, planar lipid membranes, and liposomes (203, 204). HlyA pores in erythrocytes are thought to be asymmetric, since proteolytic enzymes digest the toxin only when applied to one side of the membrane. Similarly, when monoclonal antibodies raised against HlyA were applied from the same side as the toxin, they could open or close the pore, but they could not do so when applied from the opposite side (204, 205). The physical properties of the transmembrane pore formed by E. coli hemolysin have been determined by using artificial lipid bilayers; it has a diameter of about 1.0 nm, a conductance of about 500 pS in 0.15 M KCl, and a mean lifetime of 2 s at low transmembrane voltages (21). HlyA pores in erythrocytes have been similarly sized, with a predicted physical cutoff of around 2 kDa (27). Patch-clamped human macrophages have been used as targets for HlyA, and this has confirmed that the leukotoxic action of HlyA is also due to the formation of pores with very different properties from the endogenous pores already present in the cell membrane (204). Similar physical properties have been assigned to the pores formed by other members of the HlyA toxin family (22, 55, 194, 203, 258), suggesting that pore formation is a closely conserved step in the action of these toxins.

For many bacterial toxins unrelated to E. coli HlyA, pore formation is synonymous with oligomerization, with their structures varying from trimer to 100-mer. V. cholerae El Tor cytolysin may form 3- to 5-mer oligomeric structures (206). E. coli heat-labile enterotoxin and P. aeruginosa cytotoxin form pentameric pores (228, 307), while aerolysin and S. aureus alpha-toxin oligomerize to produce heptameric pores (236, 278). The two gram-positive toxins streptolysin O and pneumolysin polymerize to form pores with a large and variable oligomeric superstructure of between 25 and 100 monomers (214, 235).

Oligomeric forms of staphylococcal alpha-toxin and streptolysin O are unusually stable and can be extracted intact from membranes with detergent (26, 90). In contrast, HlyA recovered from deoxycholate-solubilized erythrocyte membranes is recovered in a monomeric form, indicating either that oligomerization is not required for pore formation or that oligomers are dissociated in the detergent (27). The lower stability, larger size, and lower solubility of HlyA-related toxins have made elucidation of the accurate subunit stoichiometry of the membrane-inserted pore even more complex and left it currently unresolved. As with streptolysin O and staphylococcal alpha-toxin, membrane insertion of HlyA and CyaA is believed to occur through a monomolecular mechanism (24, 286), with oligomerization, if any, occurring by the subsequent addition of monomers within the membrane. It has been estimated that only one to three HlyA molecules form the pore (21, 203), while CyaA pores in artificial lipid bilayers indicate a functional unit of a trimer or larger (297). The increases in membrane conductance on exposure to HlyA and Apx toxins have been explained by an association between nonconducting monomers and conducting oligomers (21, 194). The complementation of inactive deleted variants of HlyA to produce hemolytic activity and CyaA to produce cytotoxic activity has suggested that two or more toxin molecules aggregate before pore formation and substantiate the view that oligomerization is involved (133, 189). However, in the absence of a defined pore structure, it is worth bearing in mind that membrane disruption may occur through mechanisms other than formation of discrete pores. Examples are provided by the antibacterial peptide cecropin, where a "carpet" of peptide monomers disrupts phospholipid packing (93), and complement-mediated lysis by the formation of "leaky patches" in the lipid bilayer (327). Indeed, the exposure of membranes to toxin at high concentration or for long times may lead to similar lesions through a detergent-like mechanism (210, 231).

A highly conserved region (amino acids [aa] 238 to 410) of HlyA (20, 61) is essential for lysis and is believed to be involved in pore formation (Fig. 2). It spans the only pronounced hydrophobic sequences in the otherwise hydrophilic HlyA protein, and secondary-structure predictions suggest that it comprises four membrane-spanning -helices, each of 21 aa. The sequences preceding and between these four putative HlyA -helices have strong amphipathic -helical properties and possess membrane-binding characteristics. It has been suggested that these repeated hydrophobic and amphipathic helices may insert into the cytoplasmic membrane and form a pore, allowing the influx of extracellular calcium and the escape of potassium (188, 203). Mutations altering the hydrophobicity of this region reduce or abolish the pore-forming activity of the protein on erythrocytes and artificial membranes (186-188).

View larger version (18K): [in this window] [in a new window]   FIG. 2.   Maturation of pro-HlyA to the active toxin by HlyC and acyl-ACP. The positively charged HlyC homodimer associates with the negatively charged ACP. Following binding to the HlyC recognition domains, FAI and FAII, the acyl chain is transferred to the corresponding acyl modification sites, K564 (KI) and K690 (KII). The acylated lysines KI and KII are shown relative to the hydrophobic pore-forming domain and the glycine-rich Ca2+-binding repeats of pro-HlyA.

The conserved glycine-rich repeat domain associated with HlyA Ca²⁺ binding is required for the hemolysis of erythocytes but not for pore formation in asolectin lipid bilayers, since its deletion leaves the pore-forming ability of HlyA unaffected (187) (Fig. 2). In addition, HlyA shows full channel-forming activity in artificial lipid bilayers even in the presence of 5 mM EDTA (74). This suggests that while Ca²⁺ binding is critical at some stage of the lytic process such as promoting the irreversible insertion of the toxin into the membrane (11), it does not directly contribute to the pore-forming structure (37, 187). It also suggests that the asolectin lipid bilayer system is not a true reflection of binding and pore formation in real cell membranes, a conclusion supported by the finding that pore formation in asolectin bilayers is also independent of acylation (190). The conformation of the repeat region, which remains on the same side as that from which the toxin approached the membrane, may affect the state of the pore since the binding of a monoclonal antibody to this region led to pore closure (205).

Mutations in the LPS biosynthetic genes affect the expression and/or activity of E. coli HlyA (17, 286, 314). A transposon insertion in rfaP (required for attachment of phosphate-containing substituents to the LPS inner core) reduces the extracellular activity but not the export of HlyA as a consequence of the formation of large extracellular toxin aggregates; biological activity is restored with chaotropic agents (286). Why this occurs is not clear, but LPS possesses calcium-binding sites that may act as outer membrane reservoirs of calcium, and alterations in LPS structure may reduce the availability of calcium, resulting in the secretion of an inactive, unstable HlyA prone to aggregation. Alternatively, direct contact between wild-type LPS and HlyA may prevent toxin aggregation.

None of the dramatic effects of HlyA on eukaryotic host cells outlined above occur in vivo if the toxin is produced in the absence of the cotranslated protein, HlyC. The elucidation of the reason for this strict requirement has revealed a mechanism of bacterial toxin maturation unknown outside the HlyA family.

ESSENTIAL MATURATION OF HLYA: A UNIQUE FATTY ACYLATION

Top Previous Next References

Bacterial Mechanism for Modifying Lysine Residues

HlyA toxin is synthesized as an inactive 1,024-residue protoxin, pro-HlyA, which is activated intracellularly to the mature toxin by the action of the cosynthesized HlyC. It was shown by using an in vitro system that this maturation is a fatty acylation and that HlyC is a homodimeric putative acyltransferase, using acyl-acyl carrier protein (acyl-ACP) as the fatty acid donor (112, 128, 129, 132, 282, 283) (Fig. 2). In in vitro reaction mixtures containing only purified acyl-ACP, HlyC and pro-HlyA, the acquisition of hemolytic activity is directly related to the binding of fatty acid by pro-HlyA. Mass spectrometry and Edman degradation of proteolytic products from mature HlyA toxin activated in vitro by HlyC and [³H]acyl-ACP revealed two fatty-acylated internal lysine residues, K564 (KI) and K690 (KII), and resistance of the acylation to hydroxylamine suggested that the fatty acid is amide linked. Substitution of the two lysines confirmed that they are the only sites of acylation and showed that although each is acylated in the absence of the other, both sites are required for in vivo toxin hemolytic activity (282). The K564 and K690 residues were subsequently confirmed to be modified in in vivo-synthesized and secreted HlyA by peptide mapping and two-dimensional electrophoresis (190). The internal pro-HlyA modification explains the earlier isolation of a monoclonal antibody that recognized only the active form of HlyA, specifically epitope aa627 to 726 (238).

The other HlyA-related toxins secreted by pathogenic gram-negative bacteria all require HlyC-type protoxin activation, and indeed many (PvxA, MmxA, ApxIA, and LktA) have been activated by E. coli HlyC (87, 106, 167). In addition, CyaC, ApxIIC, and AaltC are able to activate pro-LktA (175, 202, 322). CyaA, the adenylate cyclase hemolysin of B. pertussis, requires posttranslational activation both to deliver its catalytic domain into the mammalian target cell (cell-invasive activity) and to form transmembrane channels (hemolytic activity). It has been demonstrated by mass spectrometry that CyaA toxin secreted by B. pertussis is modified by amide-linked palmitoylation on the -amino group of K983 (analogous to HlyA KII, K690) (108). Only modification sites of HlyA and CyaA have been demonstrated to be acylated, although the loss of activity caused by deletion of the region between aa 358 to 548 of P. haemolytica LktA is compatible with the KI residue, K554, as a potential site of modification (61), and activation of a CyaA-LktA hybrid toxin supports the notion that the region aa 379 to 616 of LktA is important for LktC recognition (322). Substitution of LktA K554 (to T and C) reduced the lytic activity of LktA against bovine lymphocytes by only ca. 40%, but the retention of lytic activity does not rule out K554 as an LktA acylation site, since HlyA mutants with either KI or KII deleted also retained about 50% activity against BL-3 cells despite being inactive to erythrocytes (239, 282). Bovine lymphocytes may be less sensitive to the state of acylation of these toxins than are either erythrocytes or human lymphocytes, but it remains a possibility that LktA contains an acylation site different from the KI and KII sites identified for HlyA and CyaA. No data for the other toxins have been presented.

The ability to transfer an acyl group to an internal lysine residue of a target protein distinguishes HlyC from all other bacterial acyltransferases, and hlyA⁺ hlyC strains are nonhemolytic, indicating that the various constitutive acyltransferases of E. coli are unable to substitute for HlyC to even a small degree. HlyC has no significant sequence homology to known acyltransferases such as the lipid A acyltransferases (3, 54, 56, 158), the Rhizobium Nod factor acyltransferases NodL and NodA (35, 69), or the well characterized acyltransferases such as glycerol-3-phosphate acyltransferases (182) or eukaryotic N-myristoyltransferases (140, 332). HlyC may therefore be an acyltransferase that is structurally and functionally distinct from all other acyltransferases.

Little is known about the biochemical properties of HlyC, the enzyme responsible for protoxin acylation. Bacterial and eukaryote acyltransferases generally accept either acyl coenzyme A (acyl-CoA) or acyl-ACP as an acyl donor, but ACP is a strict requirement for HlyC-directed pro-HlyA acylation (acyl-CoA cannot be used). Myristoyl-ACP gave the highest hemolytic activity of HlyA acylated in vitro with a range of fatty acids (C₁₂ to C_18:1) (132), which is consistent with an apparent selection of myristoylation at both the KI and KII sites in vitro. The affinity of KI for myristoyl-ACP (a saturated acyl group of 14 carbons) in vitro is approximately twice that for palmitoyl-ACP (a saturated acyl group of 16 carbons), while the affinity of KII for myristoyl-ACP is approximately five times that for palmitoyl-ACP (285). In vivo, both K564 and K690 appear fully acylated, suggesting that unlike CyaA, HlyA is predominantly myristoylated (190). Not only is HlyC able to discriminate between acyl-ACPs carrying fatty acids of different lengths but also it is presumably able to do this at low substrate concentrations, since E. coli does not have significant pools of acyl-ACPs with fatty acids longer than 3 carbons (249). In vitro measurements of K_m for both ACP and protoxin substrates suggest that HlyC does indeed recognize its substrates at extremely low concentrations (~100 and ~10 nM, respectively). HlyC is able to bind tightly but, it seems, noncovalently both the acyl chain from acyl-ACP and phosphopantetheine, but the locations of the active site of HlyC and regions that bind its acyl-ACP and pro-HlyA substrates are not known (284). The specificity for acyl-ACP as the acyl donor is shared in E. coli by the acyltransferases involved in lipid A biosynthesis, LpxA, LpxD, HtrB, and MsbB, but the use of a protein as a substrate is unique to HlyC (43, 158, 246). The cytosolic enzyme, LpxA, catalyzes the first step of lipid A biosynthesis, transferring (R)-3-hydroxymyristate from ACP to UDP-N-acetylglucosamine (3). LpxD is also specific for (R)-3-hydroxymyristoyl-ACP (158). The so-called "late" acyltransferases, HtrB and MsbB, can both use lauroyl-ACP and myristoyl-ACP as substrates for generating acyloxyacyl residues in lipid A, although HtrB shows a fivefold preference for lauroyl-ACP (54).

By using deleted protoxin variants and protoxin peptides as substrates in an in vitro maturation reaction dependent on only HlyC and acyl-ACP, two independent HlyC recognition domains (FAI and FAII) have been identified on the HlyA protoxin, each of which spans one of the target lysine residues, KI or KII, respectively (283) (Fig. 2). Each domain requires 15 to 30 aa for basal (ca. 10%) recognition and 50 to 80 aa for full wild-type acylation. The peptide recognition sequences for HlyC appear to be larger than those of other acyltransferases, since acylation of mutagenized target substrate proteins, chimeric proteins, and substrate peptides has indicated that peptide sequences of 4 to 15 aa are recognized by N-myristoyltransferases and prenyltransferases (198, 240, 302, 303, 331). However, these modification sites are generally at the N and C termini of the substrate proteins, where they may be relatively open and accessible. In contrast, the internal acylation sites of HlyA may require HlyC to recognize a larger topology rather than a linear sequence. While the HlyA FAI domain is symmetrically centered on the modified lysine residue, the FAII domain appears to be asymmetrical, not requiring amino acids N-terminal to the modified residue (283). The two domains are nevertheless functionally indistinguishable and compete with each other for HlyC both in cis and in trans. No other HlyA sequences are required for toxin maturation, including the immediately C-terminal Ca²⁺-binding repeats (aa 722 to 849). Indeed, in vitro, Ca²⁺ ions prevent acylation at both the KI and KII sites (283). The extreme sensitivity of the pro-HlyA activation reaction to free Ca²⁺ supports the view that intracellular Ca²⁺ levels in E. coli are too low to affect toxin activity (92) and that Ca²⁺ binding does not occur until the toxin is outside the cell.

The HlyC recognition sites of E. coli HlyA have been compared with each other and with the corresponding sequences of other members of the toxin family. Although FAI and FAII of HlyA have the same function, there is little primary sequence identity (only 21%), and this lack of homology is also evident within each site among the toxins; only 20% of the FAI residues and 17% of the FAII residues are identical in the five toxins shown (Fig. 3). This divergence may reflect differences in the HlyC-type proteins, the pattern of acylation at the two sites (where present, some toxins of the Pasteurellaceae family lack the KII target lysine), and the preferred target cell range of each toxin. Consensus sequences constructed for each of the two acylation sites have little in common apart from the central "GK" motif, suggesting that similarity between FAI and FAII may be of a higher structural order. Secondary-structure predictions of FAI and FAII indicate that both regions are rich in -turns, quite regularly spaced approximately every 10 aa, especially in FAI. Other predicted features are not shared by FAI and FAII, except perhaps for a helical region immediately N-terminal to both KI and KII. However, helical-wheel projections do not show amphipathic distributions of charged or hydrophobic residues as related to membrane-associated structures. The lack of identity between FAI and FAII domains in pro-HlyA and corresponding sites on related protoxins currently deters an explanation for the basis of HlyA recognition by HlyC.

View larger version (38K): [in this window] [in a new window]   FIG. 3.   Alignment of the E. coli HlyA FAI and FAII HlyC recognition domains with corresponding sequences of the related toxins A. actinomycetemcomitans AaltA, A. pleuropneumoniae ApxIA, P. haemolytica LktA, and B. pertussis CyaA. Identical residues are in boldface type; amino acids belonging to the same class are in capitals and non-homologous amino acids are in lowercase. Hyphens represent breaks introduced to maximize identity. Consensus sequences for FAI and FAII indicate residues shared by four toxins (boldface) and by all five toxins (underlined). A cumulative histogram representation of secondary structure predictions by the Chou and Fasman and the Robson and Garnier methods are shown under the primary sequences.

A lack of cross-complementation between protoxin activator C proteins supports the view that C-proteins may be unable to modify a KI or KII site even when it is present. E. coli HlyC is able to activate P. haemolytica LktA, whereas LktC is unable to activate HlyA and CyaA (87, 322). One explanation would be that LktC can acylate only KI even when KII is present. However, this explanation could not extend to A. pleuropneumoniae ApxIIC, which can activate P. haemolytica LktA, whereas LktC is able to activate ApxIIA only marginally, despite both toxins apparently possessing only KI (202) (Table 3). It seems possible that there are additional factors influencing the recognition between protoxin and HlyC-like activator proteins. One such additional factor appears to be the host cell in which the toxin operon is expressed. In contrast to the native CyaA protein from B. pertussis, which is palmitoylated exclusively at K983 (KII), recombinant pro-CyaA protoxin modified by CyaC in E. coli is acylated at K983 and contains an additional acylation site at K860 (analogous to KI) (109). It is perhaps this anomalous acylation in E. coli that allows CyaC to activate LktA (322). The specificity of CyaC peptide recognition may be affected by the host cell ACP, but this presumes that it is a C-ACP complex which recognizes the toxin. Such a mechanism is supported for HlyC by in vitro kinetic data that also indicate the involvement of a ternary HlyA-HlyC-ACP complex during the transfer of fatty acid from acyl-ACP to pro-HlyA (284). An ordered Bi-Bi mechanism would be analogous to that of N-myristoyltransferases, for which it is the binding of acyl-CoA which determines subsequent peptide binding (252, 332). This might explain the earlier failures of LktC to activate the proforms of HlyA, ApxIIA, and CyaA (87, 202, 322) in experiments which were performed in or with extracts from an E. coli background. Alternatively, E. coli may possess an escort protein, similar to that observed with prenylated proteins, which enhances the acylation at KI but which is absent in B. pertussis (269). However, one might expect such an escort protein to have been identified by mutagenesis studies carried out in many laboratories that have successfully identified other accessory proteins, such as TolC and RfaH, required for toxin activity. Although the data available at present are incomplete and address only HlyA and CyaA, it appears that there is selection for the fatty acid attached to the toxin and that E. coli and B. pertussis have evolved hemolysin acyltransferases with different affinities for fatty acid. The E. coli HlyC protein may preferentially acylate HlyA KII with myristic acid, whereas the B. pertussis CyaC protein preferentially acylates CyaA KII with palmitic acid. Similar variations in substrate specificity between related proteins of different species appear between E. coli, Neisseria meningitidis, and P. aeruginosa LpxA, which are specific for (R)-3-hydroxymyristoyl-ACP, 3-hydroxylauroyl-ACP, and 3-hydroxydecanoyl-ACP, respectively (75, 227), and E. coli and Haemophilus influenzae HtrB, which are specific for lauroyl-ACP and myristoyl-ACP, respectively (54, 224). Not only do differential fatty acid affinities require an explanation, but so do cross-species complementation tests with C-proteins and host-specific acylation patterns. Presumably, these explanations will lie in the differences between the structures of the family toxins, activator C-protein acyltransferases, and ACPs.

An insight into the modification of HlyA might be expected from an examination of other examples of protein lipidation. We present an extensive review of these and show that while the different types of modification vary considerably in terms of the location and structure of the lipid moiety, it is possible to define universal themes in terms of their function. Despite the absence of data for acylation events exactly equivalent to that of the prokaryotic toxins, it is still possible to draw conclusions from the more extensively studied examples that are relevant to the roles that toxin acylation may play in determining biological function. Lipidated proteins found in and on eukaryotic membranes are included in this comparison since the eukaryotic cell membrane is the site of action of HlyA and its related toxins. It is also becoming apparent that these acylated proteins are the sites of action of many bacterial effector proteins either directly by modification (e.g., ADP-ribosylation, glucosylation, and deamidation of small G-proteins and heterotrimeric G-protein subunits) or indirectly by interfering with their enzymatic pathways (e.g., protein phosphatases and kinases) (84, 85).

Lipidation is involved in the maturation of many proteins in both prokaryotic and eukaryotic cells, including viral oncogene products, but it is achieved by various mechanisms which differ according to the fatty acid transferred, the amino acid modified, and the fatty acyl donor. Myristic and palmitic acids are the most common fatty acids cross-linked to proteins. Proteins sorted to the bacterial outer membrane or eukaryotic plasma membrane undergo processing in which an acyl group is attached to the N-terminal amino acid; enzymes with acyltransferase, lipase, or esterase activity use catalytic mechanisms involving ester-linked acyl groups attached to serine and cysteine residues; and eukaryotic proteins use ester-linked palmitoylation and ether-linked prenylation of cysteine residues for membrane sorting and protein-protein interaction. As can be seen below, the acylation of pro-HlyA does not equate to any of these but instead appears to be limited to its related toxins and perhaps a handful of eukaryotic proteins (Fig. 4).

View larger version (26K): [in this window] [in a new window]   FIG. 4.   Major classes of lipidated proteins in prokaryotes and eukaryotes. The structure of the lipid and their attachment to the protein peptide backbone are shown, except for the complex indirect linkages of pantetheinylated and glypiated proteins. For heterotrimeric G-proteins, the particular modified subunit is indicated in parentheses. The placement of the methyl group of geranylated proteins in parentheses indicates that carboxymethylation is not universal.

Bacterial phosphopantetheinylated proteins. The inclusion of a phosphopantetheine cofactor increases the number of thiol groups on a protein and introduces a tether to increase the flexibility of acyl chains between various active sites. As one of the most abundant proteins in E. coli, constituting 0.25% of the total soluble protein, ACP is an important example of a protein carrying an indirectly linked acyl group. In addition, the central involvement of ACP in the acylation of bacterial toxins warrants an extended account of this small but multifaceted protein. ACP is required throughout fatty acid biosynthesis, carrying fatty acids as thioester intermediates attached to the terminal sulfhydryl of a 4'-phosphopantetheine group (4'-PP), which is in turn attached to a serine residue (Ser36) via a phosphodiester linkage (192).

Eukaryotic glypiated proteins. Fatty acyl groups are linked indirectly to many eukaryotic proteins by the GPI moiety (Fig. 5). This contains an entire phospholipid, in which the lipid moiety is variable, possessing C₁₄ to C₂₄ acyl groups, and is associated with sugars and ethanolamine (77). In many instances, the inositol ring contains an additional lipid modification in the form of an ester-linked palmitic acid. The complete GPI anchor is transferred to proteins by GPI-transamidases at a specific C-terminal recognition sequence which is cleaved in the process. The GPI signal sequence is extremely degenerate but commonly features a run of 12 to 20 hydrophobic residues. GPI-anchored proteins are abundant on the cell surfaces of lower eukaryotes such as protozoa and yeasts and have a wide variety of functions such as nutrient uptake and membrane-signalling events; they include hydrolytic enzymes, receptors, cell adhesion molecules, complement inhibitors, and antigens of unknown function (77, 79, 98).

View larger version (33K): [in this window] [in a new window]   FIG. 5.   Membrane localisation of lipidated proteins on the plasma membrane of mammalian cells targeted by HlyA. Typical proteins from the major types involved in cell signalling are represented. The lipid groups are shown inserted directly into the bilayer, although other protein interactions may be involved. Many G-protein coupled receptors possess internal cysteine residues that may act as palmitoylation sites as shown, and some have been shown to be modified. The heterotrimeric G-protein  subunit, Ras and NRTK proteins shown on the inner face of the plasma membrane possess two lipids, although other isotypes may have only one. The G-protein -subunit and Src-related NRTK contain both a myristoyl group linked to the N terminus and a thioester-linked palmitoyl group. The G-protein  subunit contains a C-terminal thioether-linked geranylgeranyl isoprenoid. The Ras-related small G-protein contains a C-terminal thioether-linked farnesyl isoprenoid and a thioester-linked palmitoyl group. The structure of a typical GPI-anchored protein with its constituent inositol, glucosamine, mannose, and ethanolamine groups is shown. An additional palmitoyl chain may be attached to the inositol ring.

Ester-linked acylation. The attachment of lipid groups to proteins through ester bonds provides a labile connection suitable for a transient role such as in catalytic mechanisms. Some enzymes in the biosynthetic pathways of fatty acids, phospholipids, and LPS bind acyl groups tightly but do not necessarily form covalent acyl-enzyme intermediates, e.g., E. coli acyl-ACP synthase (136) and -hydroxy-decanoyl-ACP dehydrase (180). However, other enzymes, particularly acyltransferases but also lipases and esterases, do use mechanisms involving oxy- and thioester acyl-enzyme intermediates (45, 66, 80, 138, 179, 181, 192, 193). In addition to these enzymes, a class of acylated proteins found in eukaryotes but apparently absent in prokaryotes carry ester-linked acyl groups not for a catalytic purpose but for structural reasons. The nature of this modification is outlined below, and its influence on protein structure and function is examined for lessons that may be applied to the acylation of HlyA.

Ether-linked prenylation. Several eukaryotic intracellular proteins are post-translationally modified by farnesyl (C₁₅) or, more commonly, geranylgeranyl (C₂₀) unsaturated isoprenoid lipids, attached through thioether bonds to cysteine residues at or near their C terminus (Fig. 5). This constitutive process results in a stably modified protein. Three protein prenyltransferases are responsible for the modification of separate substrate groups. Farnesyltransferase and geranylgeranyltransferase I recognize proteins with a -CaaX C-terminal motif. Following prenylation, the three C-terminal residues are removed from most CaaX-type proteins and the prenylcysteine residues is methylated at its exposed carboxyl group. Geranylgeranyltransferase II recognizes C-terminal double cysteine motifs such as -CC or -CxC (198, 331).

(i) N-terminally acylated bacterial lipoproteins. Lipoproteins exported to the periplasm or outer membrane of gram-negative bacteria are modified with diacylglycerol, via thioether linkage to cysteine, and with an amide-linked fatty acid (65% palmitate) on the N terminus (116). The precursor protein undergoes three reactions directed by enzymes in the inner membrane (i.e., phosphatidylglycerol diacylglyceryl transferase, signal peptidase II, and apolipoprotein N-acyl transferase), resulting in the generation of N-acyl diacylglycerylcysteine as the N-terminal amino acid (254). The first known example of bacterial lipoprotein, E. coli murein (also called Braun's lipoprotein), occurs as a free form and a bound form covalently attached to the peptidoglycan layer by a peptide linkage via the -NH₂ group of the C-terminal lysine. The fatty acids transferred to the prolipoprotein are taken from the phospholipid pool, with the major source being phosphatidylethanolamine (104, 135). Prolipoprotein acylation produces lysophospholipids that are reacylated by the inner membrane 2-acyl-glycerophosphoethanolamine acyltransferase. This acyltransferase can use acyl-ACPs from fatty acid biosynthesis or can convert fatty acid to acyl-ACP in the presence of ATP-Mg²⁺ (58).

(ii) N-myristoylated eukaryotic proteins. Another N-terminal lipidation is undergone by several eukaryotic proteins in which a myristoyl group is attached to the -NH₂ group of the N-terminal glycine (position 2) following removal of the initiator methionine (140). The enzyme responsible, myristoyl-CoA:protein N-myristoyl transferase, is ubiquitous in eukaryotic cells. It was first isolated from Saccharomyces cerevisiae and has been extensively investigated (140, 302, 303, 332). It is almost exclusively specific for myristic acid and has stringent sequence requirements with respect to the 5 aa immediately distal to the glycine residue (140). An example of a myristoyl-CoA:protein N-myristoyl transferase using an acyl substrate other than myristoyl-CoA has been characterized in the retina (141, 163). Protein myristoylation is cotranslational and constitutive, and the product is usually a stably modified protein, although lipid removal from a mature protein has been reported (196).

(iii) Internal amide-linked acylation. As we have shown above, protein acylation is generally divided into labile internal modifications and stable modifications at the N and C termini. The mechanism of the stable acylation of HlyA at internal lysines via amide bonds is unique to the HlyA toxin family; no other prokaryotic proteins are known to be modified in this way. However, there are eukaryotic proteins that possibly have similarities to the HlyA acylation mechanism, but there has been little detailed characterization since their identification in the 1980s. Internal acylated lysines have been defined in four eukaryotic proteins, while another three proteins may possess hydroxylamine-resistant acyl groups not present at the N terminus. None of the associated acyltransferase activities has been characterized (Table 4).