Aggregatibacter actinomycetemcomitans Leukotoxin Utilizes a Cholesterol Recognition/Amino Acid Consensus Site for Membrane Association

Background: A repeats-in-toxin (RTX) leukotoxin and its integrin receptor aggregate in cholesterol-rich lipid rafts.

Results: The affinity of the toxin to cholesterol is driven by a cholesterol recognition/amino acid consensus (CRAC) motif.

Conclusion: Leukotoxin cytotoxicity is regulated by the CRAC motif.

Significance: Other RTX toxins contain this CRAC motif, suggesting a role for cholesterol recognition in RTX cytolysis.

Abstract

Aggregatibacter actinomycetemcomitans produces a repeats-in-toxin (RTX) leukotoxin (LtxA) that selectively kills human immune cells. Binding of LtxA to its β₂ integrin receptor (lymphocyte function-associated antigen-1 (LFA-1)) results in the clustering of the toxin·receptor complex in lipid rafts. Clustering occurs only in the presence of LFA-1 and cholesterol, and LtxA is unable to kill cells lacking either LFA-1 or cholesterol. Here, the interaction of LtxA with cholesterol was measured using surface plasmon resonance and differential scanning calorimetry. The binding of LtxA to phospholipid bilayers increased by 4 orders of magnitude in the presence of 40% cholesterol relative to the absence of cholesterol. The affinity was specific to cholesterol and required an intact secondary structure. LtxA contains two cholesterol recognition/amino acid consensus (CRAC) sites; CRAC³³⁶ (³³³LEEYSKR³³⁹) is highly conserved among RTX toxins, whereas CRAC⁵⁰³ (⁵⁰¹VDYLK⁵⁰⁵) is unique to LtxA. A peptide corresponding to CRAC³³⁶ inhibited the ability of LtxA to kill Jurkat (Jn.9) cells. Although peptides corresponding to both CRAC³³⁶ and CRAC⁵⁰³ bind cholesterol, only CRAC³³⁶ competitively inhibited LtxA binding to this sterol. A panel of full-length LtxA CRAC mutants demonstrated that an intact CRAC³³⁶ site was essential for LtxA cytotoxicity. The conservation of CRAC³³⁶ among RTX toxins suggests that this mechanism may be conserved among RTX toxins.

Introduction

Aggregatibacter actinomycetemcomitans is a pioneer colonizer of the upper aerodigestive tract of man. In the establishment of the ecological niche of the organism, it relies upon an array of virulence determinants, one of which is a leukotoxin (LtxA) that selectively kills human immune cells (1) in a process that is initiated by recognition of a cell surface molecule that leads to a multistep cascade and results in targeted cell death (2). LtxA is a member of the repeats-in-toxin (RTX)² family of cytotoxic proteins (3), which share cell type specificity that is driven by their association with β₂ integrin heterodimers, such as lymphocyte-function associated antigen-1 (LFA-1) (4) and macrophage-1 antigen (Mac-1)/complement receptor 3 (CR3) (5).

Five genes are required for successful translation and secretion of LtxA (the structural toxin gene product) in A. actinomycetemcomitans. Four of the genes, ltxCABD (in transcriptional order) are located in the leukotoxin operon. A fifth gene, tdeA (6), is not part of the ltx operon and is located 572 kb downstream of this site. Prior to secretion, Gram-negative bacterial protein toxins are translated as protoxins and must be post-translationally modified to achieve biological activity. This modification takes various forms with bacterial protein toxins. In the case of RTX toxins, the RTXC protein, an acylase, catalyzes the attachment of fatty acyl chains to internal lysine residues of the toxin (7–9) in the bacterial cytoplasm; this acylation process is necessary for the toxin to achieve its biological activity (10). The remaining three gene products (LtxB, LtxD, and TdeA) form a type I secretion system and export LtxA directly from the bacterial cytoplasm to the external environment without the need for a periplasmic intermediate.

Hydropathy analysis of the amino acid sequence of LtxA (114.5 kDa; 1055 amino acids) (35, 36) in combination with algorithms predicting secondary structure (37, 38) permit the division of LtxA into four domains: hydrophobic, central, repeat, and C-terminal domains (35, 36) (Fig. 1A). The hydrophobic domain (residues 1–420) contains a preponderance of hydrophobic amino acids and suggests an involvement in membrane insertion. Secondary structure predictions of the region show that half of the residues form either amphipathic or hydrophobic α-helices, which are often associated with toxin-target cell membrane interactions. Comparison of the amino acid sequence of LtxA with other RTX toxins, such as Escherichia coli α-hemolysin (HlyA) and Mannheimia hemolytica LktA, show a remarkable concordance in the clustering of hydrophobic and hydrophilic residues among the three toxins in this area (35, 36). The central domain (residues 421–730) is composed largely of hydrophilic residues, and the conservation of structure among the various RTX toxins is less strong. Located within the central region are the two acylated lysine residues (Lys⁵⁶² and Lys⁶⁸⁷) that are necessary for biological activity (10) (Fig. 1A, red circles labeled A). The repeat region (residues 730–900) contains a series of 14 repeated nonapeptides with the consensus sequence of GGXG(N/D)DX(L/I/F)X (where X represents any amino acid), which together form a unique calcium-binding structure that is called a parallel β-helix or parallel β-roll (39, 40). LtxA has been shown to bind calcium as well (41). Finally the C-terminal domain (positions 901–1055) is required for secretion (42).

FIGURE 1.

Structure and purity of LtxA. A, the sequence of LtxA contains an N-terminal hydrophobic domain, a central domain that contains two internal lysines (red circle labeled A) that are the sites of acylation, a repeat domain, and a C-terminal domain. The CRAC sites identified in this work are located near the hydrophobic region (yellow circle labeled C). B, SDS-PAGE gel and Western blot of LtxA^wt, LtxA^CS001, and LtxA^SM003.

The RTX toxins are known to be secreted into the aqueous solution but also to be membrane-interacting proteins (11). Neither the water-soluble nor the membrane-embedded structures of these toxins (and LtxA in particular) have been solved; however, it has been proposed that this family of toxins undergoes a conformational change upon membrane association (12). We have recently shown that LtxA undergoes distinct lipid-specific secondary structure changes upon binding to or disruption of lipid bilayers (13).

LtxA and other RTX toxins bind to various β2 integrins and cluster in lipid rafts (14–16). The result of this interaction mimics aspects of the integrin (17) activation signal, including the elevation of cytosolic Ca²⁺, activation of calpain, and cleavage of talin. The cleavage of talin frees LFA-1 from the cytoskeleton and allows lateral movement of the LFA-1·LtxA complex to lipid rafts (14). Although talin cleavage and release of LFA-1 is the initial step in toxin/integrin-driven raft clustering, the process that moves the LtxA·LFA-1 complex to the raft and subsequently maintains them in the raft is not clear. This observation has led us to hypothesize that LtxA has an affinity for cholesterol, a primary raft component, and this interaction may be responsible for the LtxA·LFA-1 clustering observed in intoxicated cells.

EXPERIMENTAL PROCEDURES

Chemicals

NaCl, CaCl₂, NaN₃, HEPES, cholesterol, guanidine hydrochloride (GuHCl), and KH₂PO₄ were purchased from Sigma. Lipids, including 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1-stearoyl-2-oleoyl-sn-glycero-3-phosphocholine (SOPC), and 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), were purchased from Avanti Polar Lipids (Alabaster, AL). Ergosterol was purchased from MP Biomedicals (Solon, OH). Peptides were synthesized and purified (98% purity) by Biomatik (Wilmington, DE).

Bacterial Strains, Growth Conditions, and Plasmids

Three A. actinomycetemcomitans strains were used in this work, JP2, CS001, and SM003 (18). JP2 produces a wild type LtxA (LtxA^wt), CS001 lacks the ltxC gene and therefore produces an unacylated LtxA (LtxA^CS001), and SM003 has been complemented with the ltxC gene in trans and produces an acylated LtxA (LtxA^SM003). The A. actinomycetemcomitans strains were grown overnight in AAGM broth (19) supplemented with antibiotics (JP2: 12.5 μg/ml vancomycin and 75 μg/ml bacitracin, CS001: 40 μg/ml kanamycin, SM003: 40 μg/ml kanamycin and 2 μg/ml chloramphenicol). LtxA was purified as described previously (20). The purity of the toxin was confirmed by Coomassie staining of SDS-PAGE and the specificity by cross-reactivity with anti-LtxA antibody (Western blot). We confirmed the activity using a cytotoxicity assay. The purity of the three proteins is shown in Fig. 1B, and the cytotoxicity is shown in Table 1.

TABLE 1.

Cytotoxicity of LtxA and Acylation Mutants

LtxA^wt, LtxA^renatured, LtxA^CS001 (nonacylated), and LtxA^SM003 (acylated) were incubated with Jn.9 cells. Cytotoxicity was measured using a trypan blue assay and was calculated as the number of cells killed relative to an untreated control. Each value represents the mean of three experiments ± standard deviation.

Protein	Cytotoxicity
	% untreated control
LtxAwt	0.77 ± 0.08
LtxArenatured	0.20 ± 0.15
LtxACS001	0.31 ± 0.24
LtxASM003	0.54 ± 0.08

Production of CRAC mutant LtxA was accomplished in E. coli using a pSHH plasmid, a pUC19-based plasmid containing the ltxA promoter, ltxC, and ltxA (21). Active LtxA is constitutively expressed in the cytosol.

The CRAC mutants were constructed by substituting proline for tyrosine at amino acid positions 336 (for CRAC^Y336P) or 503 (for CRAC^Y503P) of ltxA. Site-directed mutagenesis was performed using a QuikChange® site-directed mutagenesis kit (Agilent Technologies, Inc., Santa Clara, CA), according to the manufacturer's instructions. The primers containing substitutions were designed using the OligoPerfect^TM Designer as shown in Table 2. The reactions were performed on an automated thermal cycler with an initial step of 30 s at 95 °C, followed by PCR amplification for 16 cycles of 30 s at 95 °C, 1 min at 55 °C, and 3 min at 68 °C. The obtained PCR products were transformed into DH5α-T1 cells, and the mutant clones were selected on LB agar plates with 50 μg/ml of ampicillin.

TABLE 2.

Primers used for site-directed mutagenesis of ltxA

The substituted nucleotides are in bold type.

Primer	Sequence	Mutation site
ltxA_1012_FW	5′-AGAATGCTTGAGGAACCCTCGAAACGCTTTAA-3′	CRAC336
ltxA_1012_RV	5′-TTAAAGCGTTTCGAGGGTTCCTCAAGCATTCT-3′	CRAC336
ltxA_1513_FW	5′-AAAGCTTATGTGGATCCTTTGAAAAAGGGTGA-3′	CRAC503
ltxA_1513_RV	5′-TCACCCTTTTTCAAAGGATCCACATAAGCTTT-3′	CRAC503

CRAC^Y336P and CRAC^Y503P were produced from E. coli using the following procedure. Overnight cultures of E. coli DH5α-T1 containing the plasmid were used at a 1:50 dilution to inoculate 200 ml of LB/ampicillin (50 μg/ml), which was grown to an A₆₀₀ of ∼0.4. The cultures were centrifuged and resuspended in 6 ml of buffer (20 mm Tris-HCl, 250 mm NaCl, 0.2 mm CaCl₂, pH 6.8), sonicated (six times for 45 s, on ice), and centrifuged to remove cell debris (12,000 × g, 15 min, 4 °C).

LtxA (LtxA^renatured) was denatured by incubating 1 μm LtxA^wt with an equal volume of 16 m GuHCl overnight (12). Following incubation, the GuHCl was exchanged with liposome buffer (150 mm NaCl, 5 mm CaCl₂, 5 mm HEPES, and 3 mm NaN₃, pH 7.4) for use in the SPR experiments.

Cell Culture

Jn.9, a subclone of the Jurkat cell line (22), a gift from Dr. Lloyd Klickstein (Novartis Institute for Biomedical Research, Cambridge, MA), was maintained as described previously (14) with 2 mm l-glutamine and 50 μg/ml gentamicin.

Liposome Preparation

Lipids dissolved in chloroform were added to a glass vial in the required amounts (24). The chloroform was evaporated under a stream of nitrogen, and the residual chloroform was removed under vacuum, creating a thin lipid film on the glass surface. Liposomes were created by hydrating the lipid film with liposome buffer (150 mm NaCl, 5 mm CaCl₂, 5 mm HEPES, 3 mm NaN₃, pH 7.4). Large unilamellar vesicles were formed by extruding the liposome solutions through a 200-nm polycarbonate membrane (24, 25).

The lipid compositions used in this work were chosen based on their raft-like nature (26). The liposomes were composed of DMPC with 0, 20, 40, or 60% cholesterol (or ergosterol). At room temperature, DMPC with 0% cholesterol exists in a non-raft-like state (liquid disordered); at 40 and 60% cholesterol, DMPC exists in a raft-like state (liquid ordered); and at 20% cholesterol, both phases coexist (27, 28).

Circular Dichroism Spectroscopy

The extent of unfolding of LtxA^renatured was determined using CD spectroscopy. Three LtxA samples were prepared at a concentration of 7 μm. 1) Native LtxA^wt was suspended in liposome buffer, 2) LtxA^denatured was prepared as described above (12) and left in GuHCl, and 3) LtxA^renatured was prepared as described above, and the buffer was exchanged with liposome buffer (to recreate the conditions of the SPR experiment).

CD spectroscopy analysis was performed on a Jasco J-810 spectropolarimeter, as discussed in our previous studies (13). The samples were analyzed at 25 °C in a 0.2-mm-path length quartz cell using the step-scanning mode from 260 to 200 nm, with a 1-nm wavelength step, 10-s averaging time, and 1-nm bandwidth. The spectra were recorded under identical conditions for an n = 3 for each sample.

Surface Plasmon Resonance

Liposomes were tethered to an L1 Biacore chip by flowing them over the chip at a rate of 5 μl/min, until a mass corresponding to ∼1000 response units (∼5 μl) had been added to the surface (30). A small volume of NaOH (5 μl) was added to remove unbound liposomes.

For kinetic analysis, the flow rate was increased to 30 μl/min before the toxin injection was initiated. Each toxin injection was 60 μl, followed by a 180-μl dissociation of buffer only. The surface was regenerated with an injection of 65 μl of 0.5% SDS. The 10 toxin concentrations ranged from 0 to 500 nm in a 1:2 dilution series. All SPR measurements were performed on a Biacore® 3000, and the data were evaluated using the BIAevaluation® software. The data were fit using a 1:1 Langmuir binding model to obtain the equilibrium dissociation constant (K_D), as well as the association (k_a) and dissociation (k_d) rates, where K_D is given by the following equation.

For steady state competition analysis, liposomes containing 40% cholesterol were tethered to an L1 chip. The flow rate was set at 5 μl/min, and 50 μl of toxin was injected, followed by a 25-μl dissociation of buffer only. The surface was regenerated with SDS, as described above. In the CRAC peptide competition experiments, the toxin, at a concentration of 1000 nm, was incubated with an equal volume of peptide at concentrations of 0, 500, 1000, 2000, and 4000 nm (molar LtxA:peptide ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4). The inhibition was defined as follows,

where R_LtxA is the response units of the LtxA-only run, and RU_LtxA/peptide is the response units of each run containing LtxA and peptide. Response units represent the surface coverage. The response units were measured at the end of the injection, just before the dissociation was initiated, at the same time point for each experiment. In the comparison of binding to cholesterol-containing liposomes by the acylation mutant toxins, LtxA^CS001 and LtxA^SM003, at concentrations of 250 nm, were injected over liposomes containing 40% cholesterol that had been tethered to the L1 chip.

Differential Scanning Calorimetry

The DSC experiments were performed as described previously (31), with the following modifications. Approximately 1 mg of the dried peptide was resuspended in 100 μl of methanol and added to a solution of SOPC with or without cholesterol in chloroform:methanol (2:1) to give the desired final ratio of peptide to lipids. The solvent was removed under nitrogen gas, and the final traces of solvent were removed in a vacuum desiccator. The films were hydrated with PIPES buffer (20 mm PIPES, 140 mm NaCl, 1 mm EDTA, adjusted to pH 7.4) and vortexed extensively. The suspension was degassed and loaded into the sample cell of a Nano II differential scanning calorimeter (Calorimeter Sciences Corp., Lindon, UT). Buffer was placed in the reference cell. Successive heating and cooling scans were run between 0 and 45 °C, at a scan rate of 1°/min.

Cytotoxicity

A trypan blue cytotoxicity assay was used to measure cell viability (32). The percentage of cells killed by each treatment was calculated by the following,

where #_t=0 is the number of cells before treatment, and #_t=5 is the number of cells after treatment. Each %kill value was normalized to the %kill value of LtxA alone.

The effect of the CRAC peptides on LtxA-mediated cell death was measured in cytotoxicity assays, as described previously (14). Briefly, Jn.9 cells were incubated with protein samples containing (a) LtxA (1 × 10⁻⁸ m), (b) LtxA (1 × 10⁻⁸ m) + peptide (1 × 10⁻⁸ m), or (c) peptide (1 × 10⁻⁸ m). The samples were incubated at 37 °C for 5–24 h.

In the case of the full-length LtxA CRAC mutants, 50 μg of total protein in the E. coli cytosolic fraction was added to 0.5 × 10⁶ Jn.9 cells and incubated for 24 h. Cell death was measured with a trypan blue assay using a Vi-cell machine (Beckman Coulter, Hialeah, FL). Western blot analysis was used to confirm LtxA expression. An E. coli DH5α-T1 cytosolic fraction that did not contain pSHH served as a control.

Statistical Analysis

The statistical analyses were performed using either Student's t test or one-way analysis of variance using SigmaPlot^® (Systat Software, Inc. Chicago, IL). The following statistical criteria were applied: p < 0.001, p < 0.05, and p < 0.01.

RESULTS

LtxA Has a Strong Affinity for Cholesterol

Fig. 2 shows the affinity of LtxA for liposomes containing DMPC and varying amounts of cholesterol (reported as mol %). Using SPR, we measured the binding of LtxA to membranes containing 0, 20, 40, or 60% cholesterol as a function of time (Fig. 2, A, B, C, and D, respectively).

FIGURE 2.

Affinity of LtxA^wt for sterol-containing membranes. Liposomes containing 0, 20, 40, or 60% sterol were tethered to an L1 Biacore chip, and LtxA^wt was injected over the top. A–D, The sensorgrams show LtxA^wt binding to membranes containing 0% cholesterol (A), 20% cholesterol (B), 40% cholesterol (C), and 60% cholesterol (D). E, the binding affinity (K_D) of LtxA^wt for cholesterol was calculated from these sensorgrams. A one-way analysis of variance test demonstrated that the data in E are significant (p < 0.05). F–I, a second series of sensorgrams shows the binding of LtxA to membranes containing 0% ergosterol (F), 20% ergosterol (G), 40% ergosterol (H), and 60% ergosterol (I). J, the binding affinity (K_D) of LtxA for ergosterol was calculated from these sensorgrams. A one-way analysis of variance test indicates that there is no significant difference in the data (p = 0.097). K and L, the third set of sensorgrams shows the binding of LtxA^wt to membranes containing DMPC/40% cholesterol (K) and DOPC/40% cholesterol (L). M, the affinity of LtxA^wt for liposomes containing 40% cholesterol and either the saturated lipid, DMPC, or the unsaturated lipid, DOPC, is not significantly different (p = 0.207). These two lipid systems contain the same amount of cholesterol but differ in their raft-like nature, with DMPC/40% cholesterol existing in a raft-like state and DOPC/40% cholesterol existing in a non-raft-like state. The data in E, J, and M represent the averages of three independent experiments, and the error bars represent the standard deviation. The data in A–D, F–I, and K–L are representative plots.

These SPR sensorgrams were used to calculate the kinetic parameters of binding. The affinity of LtxA for the membranes increased (K_D decreased) as the amount of cholesterol increased up to 40% and then decreased (K_D increased) in membranes containing 60% cholesterol (Fig. 2E). The maximal affinity (minimal K_D) to membranes containing 40% cholesterol was approximately 4 orders of magnitude greater (10⁻¹² m) than the affinity for cholesterol-free membranes (10⁻⁸ m).

Table 3 shows the corresponding association (k_a) and dissociation (k_d) rates for the binding of LtxA to liposomes containing 0–60% cholesterol. There was no significant change in the association rate as the amount of cholesterol increased; however, the dissociation rate decreased as the amount of cholesterol in the membranes increased to 40% and then increased at 60% cholesterol. Therefore, the affinity of LtxA to cholesterol, as measured by the K_D, is almost entirely due to cholesterol-dependent changes in the dissociation rate rather than changes in the association rate.

TABLE 3.

Kinetic parameters of LtxA^wt and LtxA^renatured binding to liposomes containing varying cholesterol compositions

LtxA^wt or LtxA^renatured was injected over liposomes composed of varying amounts of cholesterol tethered to a L1 chip. The rates of association and dissociation (k_a and k_d, respectively) were fit to a 1:1 Langmuir binding model, using the BIAevaluation software. The values represent the means of four independent experiments ± standard deviation.

	ka	kd
LtxAwt	m−1s−1	s−1
0% cholesterol	7.68 ± 4.38 × 104	7.51 ± 3.95 × 10−4
20% cholesterol	7.32 ± 1.63 × 104	3.10 ± 2.26 × 10−4
40% cholesterol	6.05 ± 0.88 × 104	3.00 ± 2.59 × 10−7
60% cholesterol	6.31 ± 1.43 × 104	3.10 ± 3.09 × 10−5
LtxArenatured
0% cholesterol	7.39 ± 3.27 × 104	1.54 ± 1.01 × 10−3
20% cholesterol	6.68 ± 2.24 × 104	1.53 ± 0.03 × 10−4
40% cholesterol	6.33 ± 3.10 × 104	5.48 ± 2.47 × 10−5
60% cholesterol	3.76 ± 1.82 × 104	9.37 ± 3.83 × 10−5

Binding Is Specific for Cholesterol

We then compared the affinity of LtxA for liposomes composed of DMPC and varying amounts of the sterol ergosterol (substituted for cholesterol) as a specificity control. The structures of these two sterols are similar, with ergosterol containing an additional double bond in the B ring and a double bond on carbon 22. The SPR sensorgrams showing LtxA association to and dissociation from membranes containing 0, 20, 40, and 60% ergosterol are shown in Fig. 2 (F, G, H, and I, respectively).

As shown in Fig. 2J, the K_D values remained essentially constant as the ergosterol composition increased, with no strong affinity observed at high sterol compositions. In Fig. 2, the difference between panels E and J indicates that the strong affinity of LtxA to sterol-containing membranes is specific for cholesterol. Membranes composed of DMPC and cholesterol are known to phase separate into cholesterol-poor (liquid-disordered) and cholesterol-rich (liquid-ordered) regions, depending on the temperature and cholesterol composition of the system (28). The liquid-ordered phase is often used as a model of lipid rafts because of their similar compositions (rich in cholesterol and saturated lipids, poor in unsaturated lipids).

To determine whether the affinity of LtxA for cholesterol-containing membranes is determined by the cholesterol composition or by the raft-like nature of the membrane, we compared the affinity of LtxA for membranes composed of 60% DOPC (unsaturated)/40% cholesterol, a lipid composition that is liquid-disordered (non-raft-like), with membranes composed of 60% DMPC (saturated)/40% cholesterol, a lipid composition that is liquid-ordered (raft-like). The SPR sensorgrams showing LtxA association to and dissociation from membranes containing DMPC/40% cholesterol or DOPC/40% cholesterol are shown in Fig. 2 (K and L, respectively). The affinity of LtxA for membranes containing 40% cholesterol was not significantly different for the two lipid types (Fig. 2 M), demonstrating that it is the presence of cholesterol that regulates this strong affinity rather than the raft-like nature of the membrane.

To determine whether a specific LtxA structural element regulates the affinity of the toxin for cholesterol, we studied the binding of renatured LtxA (LtxA^renatured) to DMPC membranes containing varying amounts of cholesterol. The binding of LtxA^renatured to membranes containing 0, 20, 40, or 60% cholesterol as a function of time is shown in Fig. 3 (A, B, C, and D, respectively). As shown in Fig. 3E, similarly to folded LtxA, the affinity of LtxA^renatured reached a maximum at 40% cholesterol (minimum in K_D); however, this maximum is less pronounced (by 3 orders of magnitude) than the maximum affinity of native LtxA for cholesterol-containing membranes (Fig. 2E) and is not statistically significant. Table 3 shows the corresponding association (k_a) and dissociation (k_d) rates.

FIGURE 3.

Affinity of LtxA^renatured for cholesterol. Liposomes containing 0, 20, 40, or 60% cholesterol were tethered to an L1 Biacore chip, and LtxA^renatured was injected over the top. A–D, The sensorgrams show LtxA^renatured binding to membranes containing 0% cholesterol (A), 20% cholesterol (B), 40% cholesterol (C), and 60% cholesterol (D). E, the binding affinity (K_D) of LtxA^renatured was calculated from these sensorgrams. The CD spectra of LtxA^wt (solid line), LtxA^denatured in 8 m GuHCl (dashed lines), and LtxA^renatured (dotted lines) in liposome buffer are shown in the inset of E. A one-way analysis of variance test indicates that there is no significant difference in the data (p = 0.063). The data in E represent the averages of four independent experiments, and the error bars represent the standard deviation. The data in A–D are representative plots.

The extent of LtxA unfolding with GuHCl in the LtxA^renatured sample was analyzed by CD. As shown in the inset of Fig. 3E, the CD spectra of LtxA^wt (solid lines) indicated that the protein was folded and contained some secondary structure. In contrast, the CD spectra of LtxA^denatured in 8 m GuHCl (dashed lines) shows that the GuHCl has unfolded the protein. After GuHCl exchange with liposome buffer, LtxA^renatured (dotted lines) adopted a similar, but distinct shape as LtxA^wt, indicating that the toxin remains partially unfolded. The final GuHCl concentration in the liposome buffer-exchanged sample was 2 m; previous work has shown that at this GuHCl concentration, LtxA loses its ability to kill target cells (12). However, these results suggest that the secondary structure of LtxA is not essential for its binding to cholesterol.

LtxA Affinity for Cholesterol Is Independent of Its Acylation Status

LtxA is post-translationally modified by the addition of two fatty acid chains to two internal lysines (10). For some proteins, acylation has been proposed to enhance membrane association (34). We therefore investigated whether the association of LtxA with cholesterol depends on the presence of the acyl groups by using a nonacylated mutant (LtxA^CS001) and a reconstructed mutant (LtxA^SM003). The purity and cytotoxicity of these mutants is shown in Fig. 1B and Table 1. We measured the binding of these mutant toxins to cholesterol using a steady state SPR assay. The SPR sensorgrams are shown in Fig. 4A, and the maximal response units (a measure of binding) are shown in Fig. 4B. Binding to cholesterol was enhanced in the absence of acyl chains, relative to the fully acylated protein, LtxA^SM003, indicating that acylation is not responsible for the strong affinity of LtxA for cholesterol, and it may actually inhibit binding to cholesterol.

FIGURE 4.

Effect of LtxA acylation on binding to cholesterol. Liposomes containing 40% cholesterol were tethered to an L1 Biacore chip, and acylated or nonacylated LtxA was injected over the top. A, steady state SPR binding by nonacylated toxin (LtxA^CS001, solid line) and acylated toxin (LtxA^SM003, dashed line) to liposomes containing 40% cholesterol. B, normalized response units (indicating extent of binding) at t = 2100 s for both protein types. The difference between the two data points is not significant. The data in A and B represent the averages of three independent experiments, and the error bars represent the standard deviation.

LtxA Contains CRAC Sites

The high affinity binding of LtxA to cholesterol suggests the presence of a lipid binding site(s) within the LtxA protein. We scanned the LtxA amino acid sequence of the toxin for the (L/V)X_1–5YX_1–5(R/K) CRAC site motif (43), which has been shown to be associated with cholesterol binding. We identified two putative CRAC sites within the LtxA sequence: CRAC³³⁶ (³³³LEEYSKR³³⁹) and CRAC⁵⁰³ (⁵⁰¹VDYLK⁵⁰⁵) (Fig. 1A, yellow circles labeled C). CRAC³³⁶ is located within the hydrophobic region and is highly conserved among several RTX toxins, including E. coli HlyA, M. hemolytica leukotoxin, Actinobacillus pleuropneumoniae AppA, and A. pleuropneumoniae hemolysin (Table 4). CRAC⁵⁰³ is not conserved with other RTX toxins, although E. coli HlyA has some homology in that region. Bordetella pertussis adenylate cyclase toxin (CyaA) does not share either CRAC motif with other RTX toxins, although the deduced amino acid sequence does contain five putative CRAC sites.

TABLE 4.

RTX CRAC site homology

Sequence homology of the two CRAC sites in LtxA with other RTX toxins. Each CRAC site is bold and underlined.

Toxin	Sequence
CRAC336
A. actinomycetemcomitans (LtxA)a	332MLEEYSKRFK341
E. coli (HlyA)b	320QLESYSERFK329
M. haemolytica (LktA)c	334SLESYAERFK343
A. pleuropneumoniae (AppA)d	330LIKSYSERFQ339
A. pleuropneumoniae (HlyA)e	330QLEQYSERFK339
B. pertussis (CyaA)f	634QLDKLAQESS643
CRAC503
A. actinomycetemcomitans (LtxA)a	499AYVDYLKKGE508
E. coli (HlyA)b	487AYINYLENGG496
M. haemolytica (LktA)c	493AYVDAFEEGQ502
A. pleuropneumoniae (AppA)d	497AYVDAFEEGQ506
A. pleuropneumoniae (HlyA)e	497AYVDFFEEGK506
B. pertussis (CyaA)f	801VFVDRFVQGE810

^a Sequence accession number CAA34731.

^b Sequence accession number AAC24352.

^c Sequence accession number Q9EV31.

^d Sequence accession number 123196.

^e Sequence accession number 1710793.

^f Sequence accession number 34978355.

CRAC Peptides Inhibit Cholesterol Binding by LtxA

Synthetic peptides corresponding to the CRAC³³⁶ and CRAC⁵⁰³ sequences were synthesized (Table 5). The first set of peptides (CRAC³³⁶) corresponds to the CRAC site between residues 333 and 339, and the second set (CRAC⁵⁰³) corresponds to the CRAC site between residues 501 and 505. Within each set, peptide “WT” is the wild type sequence, peptide “MUT” contains a proline in place of the central tyrosine residue (the central tyrosine has been shown to be critical for cholesterol binding (44)), and peptide “SCM” contains the scrambled CRAC sequence.

TABLE 5.

LtxA CRAC peptides

A panel of two sets of peptides was created to correspond to LtxA CRAC³³⁶ and CRAC⁵⁰³ sequences. Each set contained peptides corresponding to the wild type CRAC sequence (WT), the CRAC sequence with the Tyr replaced by Pro (MUT), and the scrambled CRAC sequence (SCM).

Peptide	Mutation	Sequence
CRAC336WT	None	NH2-FDRARMLEEYSKRFKKFGY-OH
CRAC336MUT	Tyr → Pro	NH2-FDRARMLEEPSKRFKKFGY-OH
CRAC336SCM	Scrambled	NH2-FDRARMYEKLERSFKKFGY-OH
CRAC503WT	None	NH2-QSGKAYVDYLKKGEELA-OH
CRAC503MUT	Tyr → Pro	NH2-QSGKAYVDPLKKGEELA-OH
CRAC503SCM	Scrambled	NH2-QSGKAYYKLDVKGEELA-OH

The interaction of each of these peptides with cholesterol-containing membranes was measured using DSC. The peptide-induced rearrangement of cholesterol in a membrane into cholesterol-rich domains can be followed by measuring the formation of cholesterol crystallites when the sterol passes its solubility limit. The formation of these crystallites can be readily detected with DSC through their polymorphic thermal transition, which occurs with a characteristic hysteresis, appearing at ∼35 °C in heating scans and at ∼23 °C in cooling scans at most conventional scan rates (45).

As shown in Table 6, none of the CRAC peptides caused a change in the phase transition temperature of pure SOPC in the absence of cholesterol at a peptide mol fraction of 0.15. In addition, none of the CRAC³³⁶ peptides caused any substantial change in the gel to liquid crystalline phase transition enthalpy, nor did CRAC peptide^503WT. However, peptides CRAC^503MUT and CRAC^503SCM both lowered the transition enthalpy by ∼20%. Thus, two of the non-CRAC negative control peptides, CRAC^503MUT and CRAC^503SCM, have some interaction with SOPC, corresponding to a more hydrophobic peptide. However, by the DSC criterion, all of the other peptides have little effect on the pure lipid, as shown in Fig. 5, and thus neither of the CRAC peptides, CRAC^336WT or CRAC^503WT, penetrate deeply into the bilayer in the absence of cholesterol.

TABLE 6.

Enthalpy of SOPC gel to liquid transition with 15 mol % peptide

Each of the six CRAC peptides was incubated with liposomes containing SOPC at 15 mol % peptide.

Peptide	Mutation	Tm	ΔHa
		°C	kcal/mol
None		3.7	4.00
CRAC336WT	None	3.6	4.00
CRAC336MUT	Tyr → Pro	3.6	4.18
CRAC336SCM	Scrambled	3.6	4.25
CRAC503WT	None	3.7	4.13
CRAC503MUT	Tyr → Pro	3.7	3.40
CRAC503SCM	Scrambled	3.7	3.49

^a The transition temperature (T_m) and enthalpies (ΔH) were calculated as the average from the cooling scans only (because the heating scans did not reach a stable base line at low temperatures before the peak where heating was initiated).

FIGURE 5.

DSC scans showing the gel-to-liquid crystalline transition temperatures. Liposomes composed of 100% SOPC were mixed with 15 mol % CRAC peptides. A, CRAC^336WT; B, CRAC^336MUT; C, CRAC^336SCM; D, CRAC^503WT; E, CRAC^503MUT; F, CRAC^503SCM. Positive curves are heating scans, and negative curves are cooling scans. The curves were displaced along the y axis for ease of presentation. The scan rate was 1°/min.

With mixtures of SOPC and cholesterol at a molar ratio of 6:4, 15 mol % of each of the six peptides was used. No cholesterol crystallite transitions were observed in SOPC:cholesterol 6:4 in the absence of peptide, as observed by the absence of the characteristic transitions at 35 °C on heating and at 23 °C on cooling. In the presence of 15 mol % peptide, the characteristic transitions were observed for all peptides, as shown in Table 7 and Fig. 6. The greatest enthalpies of these cholesterol crystallite transitions were observed with the two CRAC peptides, CRAC^336WT and CRAC^503WT (Table 7). These DSC results indicate that both CRAC peptides (CRAC^336WT and CRAC^503WT) interact with cholesterol more strongly than do the control peptides (CRAC^336MUT, CRAC^336SCM, CRAC^503MUT, and CRAC^503SCM).

TABLE 7.

Enthalpy of cholesterol crystallite transitions

Each of the six CRAC peptides was incubated with liposomes containing SOPC and cholesterol (6:4) at 15 mol % peptide. The formation of cholesterol crystallites was observed by the characteristic hysteresis at approximately 35 °C in heating scans and at approximately 23 °C in cooling scans.

Peptide	Mutation	ΔHheat	ΔHcool	ΔHavga
		cal/mol	cal/mol	cal/mol
CRAC336WT	None	160	140	150
CRAC336MUT	Tyr → Pro	100	80	90
CRAC336SCM	Scrambled	80	60	70
CRAC503WT	None	190	170	180
CRAC503MUT	Tyr → Pro	95	75	85
CRAC503SCM	Scrambled	60	60	60

^a The average transition enthalpies (ΔH_avg) were calculated as the averages from the heating and cooling scans.

FIGURE 6.

DSC scans showing the characteristic cholesterol crystallite transitions at 35 °C on heating and 23 °C on cooling. The lipid mixture was SOPC:cholesterol 6:4, containing 15 mol % CRAC peptide. A, CRAC^336WT; B, CRAC^336MUT; C, CRAC^336SCM; D, CRAC^503WT; E, CRAC^503MUT; F, CRAC^503SCM. Positive curves are heating scans, and negative curves are cooling scans. The curves were displaced along the y axis for ease of presentation. The scan rate was 1°/min.

To determine whether the interaction of the CRAC peptides with cholesterol inhibits binding of LtxA to cholesterol, a series of SPR competition experiments was run. In each experiment, LtxA was incubated with increasing amounts of each CRAC peptide (molar LtxA:peptide ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4), and this mixture was injected over tethered liposomes containing 40% cholesterol.

Fig. 7A demonstrates that CRAC^336WT inhibited binding of LtxA to cholesterol. The CRAC^336MUT (Fig. 7B) control peptide only slightly inhibited binding of LtxA to cholesterol, and the CRAC^336SCM (Fig. 7C) control peptide slightly enhanced binding of LtxA to cholesterol. As shown in Fig. 7D, CRAC^503WT had a minimal effect on the binding of LtxA to cholesterol. The CRAC^503MUT peptide had an even slighter inhibitory effect on binding, and CRAC^503SCM slightly increased the binding of LtxA to cholesterol, as shown in Fig. 7 (E and F, respectively). The effects of all three CRAC⁵⁰³ peptides were statistically insignificant (CRAC^503WT, p > 0.096; CRAC^503MUT, p > 0.284; CRAC^503SCM, p > 0.390). The hydrophobic nature of these six peptides may have some effect on the binding of LtxA to the membrane, but it is clear that the largest and most significant effect is the inhibition of binding to cholesterol by CRAC^336WT. Collectively, these results suggest that CRAC³³⁶ is involved in the binding of cholesterol by LtxA, but CRAC⁵⁰³ is not. The inhibition of binding by the CRAC³³⁶ and CRAC⁵⁰³ peptides is shown in Fig. 7 (G and H, respectively).

FIGURE 7.

Inhibition of LtxA^wt binding to cholesterol-containing membranes by CRAC 1 and CRAC 2 peptides. A, CRAC^336WT at ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4 inhibits binding of LtxA^wt to membranes containing 40% cholesterol. B and C, however, CRAC^336MUT (B) and CRAC^336SCM (C) at ratios of 1:0 1:0.5, 1:1, 1:2, and 1:4 do not inhibit binding of LtxA^wt to membranes containing 40% cholesterol. D, CRAC^503WT at ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4 does not inhibit binding of LtxA^wt to membranes containing 40% cholesterol. E and F, likewise, CRAC^503MUT (E) and CRAC^503SCM (F) at ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4 do not inhibit binding of LtxA^wt to membranes containing 40% cholesterol. G, the CRAC^336WT peptide inhibits LtxA^wt binding to membranes containing 40% cholesterol (black bars), but disruption of the CRAC site by changing Tyr to Pro (light gray bars, CRAC^336MUT) or by scrambling the CRAC site (dark gray bars, CRAC^336SCM) prevents inhibition. H, the CRAC⁵⁰³ peptides (CRAC^503WT, CRAC^503MUT, and CRAC^503SCM) do not inhibit LtxA^wt binding to membranes containing 40% cholesterol. I, CRAC^336WT at ratios of 1:0, 1:0.5, 1:1, 1:2, and 1:4 only slightly inhibits binding of LtxA^wt to liposomes that do not contain cholesterol. The data represent the averages of three experiments, and the error bars represent the standard deviation. **, p < 0.01; *, p < 0.05.

In addition, the specificity of the inhibition of CRAC^336WT on binding to cholesterol was demonstrated by injecting the LtxA/CRAC^336WT mixtures over liposomes containing no cholesterol (100% DMPC). Here (Fig. 7I), no inhibition of binding was observed, indicating that this peptide inhibits binding only when cholesterol is present in the membrane, and therefore, the peptide's binding to cholesterol is what drives the observed inhibition of LtxA binding.

LtxA Binding to Cholesterol Is Required for Cytotoxicity

The effect of the CRAC peptides on the LtxA-mediated cytotoxicity of Jn.9 cells was measured. Each set of cells was incubated for 5 h with either PBS, LtxA, LtxA:CRAC peptide (1:2), or CRAC peptide alone.

Table 8 shows the % cytotoxicity values, normalized to LtxA-induced cytotoxicity, which was set as 1.00. CRAC^336WT, which inhibits LtxA binding to cholesterol, almost completely eliminated LtxA-mediated cell death (% cytotoxicity = 0.06). CRAC^336MUT and CRAC^336SCM, which do not inhibit LtxA binding to cholesterol, had minimal effect on LtxA-mediated cell death (% cytotoxicity = 0.97 and 0.98, respectively). CRAC^503WT, which only slightly inhibited LtxA binding to cholesterol, had no effect on LtxA-mediated cell death (% cytotoxicity = 1.03). The ability of each peptide to inhibit binding to cholesterol by LtxA correlates with its inhibition of LtxA-mediated cytotoxicity, indicating that binding to cholesterol by the toxin is a required step in its cytolysis.

TABLE 8.

Inhibition of LtxA-mediated cytotoxicity by CRAC peptides

LtxA was preincubated with the CRAC peptides at a molar ratio of 1:2 before being incubated with Jn.9 cells.

Toxin	Peptide mutation	Cytotoxicitya
		% untreated control
None		0.00
LtxA		1.00
LtxA + CRAC336WT	None	0.06 ± 0.10
LtxA + CRAC336MUT	Tyr → Pro	0.97 ± 0.43
LtxA + CRAC336SCM	Scrambled	0.98 ± 0.37
LtxA + CRAC503WT	None	1.03 ± 0.28
CRAC336WT alone	None	0.19 ± 0.38
CRAC336MUT alone	Tyr → Pro	0.34 ± 0.25
CRAC336SCM alone	Scrambled	0.28 ± 0.36
CRAC503WT alone	None	0.00 ± 0.56

^a Cytotoxicity was measured using a trypan blue assay (32). The number of cells killed under each experimental condition was normalized to the number of cells killed by LtxA alone (after 5 h). Cytotoxicity was calculated as the number of cells killed relative to an untreated control. Each value represents the mean of three experiments ± standard deviation.

CRAC³³⁶ Is Required for LtxA to Kill Target Cells

To confirm the in vitro binding studies that utilized CRAC peptides, point mutations were induced in ltxA CRAC³³⁶ and CRAC⁵⁰³ sites to determine whether cholesterol binding affected LtxA-mediated cytolysis in vivo. Using site-directed mutagenesis, proline was substituted for tyrosine at amino acid positions 336 (for CRAC³³⁶) or 503 (for CRAC⁵⁰³) of individual ltxA. LtxA^WT and ltxA with the CRAC mutant genes were cloned into pSHH and expressed in tandem with ltxC under the control of the wild type leukotoxin promoter. Overnight cultures of LtxA and LtxA-CRAC mutants were constitutively expressed in the E. coli DH5α cytosol and stained with rabbit anti-LtxA antibody (Fig. 8, insert). Following sonication, supernatants from each experimental group and controls were normalized for protein content and incubated with Jn.9 target cells. The cytotoxicity was measured with a trypan blue cytotoxicity assay (32).

FIGURE 8.

Mutation of CRAC³³⁶ affects LtxA cytotoxicity. Point mutations were induced in ltxA by substituting proline for tyrosine at amino acid positions 336 (for CRAC³³⁶) and 503 (for CRAC⁵⁰³), using site-directed mutagenesis. LtxA wild type and CRAC mutant genes were cloned into pSHH and expressed in tandem with ltxC under the control of the native leukotoxin promoter. The inset shows a Western blot of the cytosolic fractions of wild type and CRAC mutants stained with anti-LtxA antibody. Overnight cultures of LtxA and LtxA CRAC mutants were constitutively expressed in the E. coli DH5α cytosol. Jn.9 cells were treated with 50 μg of total protein from the E. coli cytosolic fractions for 24 h. Jn.9 cells exposed to LtxA^WT sonicates expressed in E. coli had a reduced viability after 24 h of exposure (41.0 ± 15.0%) when compared with a control sonicate that did not contain ltxA (94.4 ± 0.5%, p < 0.001). A point mutation in ltxA that replaced Tyr³³⁶ with Pro³³⁶ (CRAC^Y336P) yielded a toxin mutant incapable of killing target cells (87.7 ± 3.3%, p = 0.04). On the other hand, whereas in vitro studies had demonstrated the ability of CRAC⁵⁰³ to bind cholesterol, cells exposed to the CRAC^Y503P mutant had a viability (49.7 ± 19.3%, p = 0.02) that was only slightly higher than in cells exposed to wild type toxin. A sonicate from E. coli DH5α containing the pSHH empty vector served as the negative control. Cell death was measured with a trypan blue assay using a Vi-cell machine. ***, p < 0.001; **, p < 0.05, relative to the empty vector control.

Jn.9 cells exposed to LtxA^WT sonicates expressed in E. coli had a reduced viability after 24 h of exposure (41.0 ± 15.0%), when compared with a control sonicate that did not contain ltxA (94.4 ± 0.5%, p < 0.001). Cells exposed to the CRAC^Y503P mutant had a viability (49.7 ± 19.3%, p = 0.02) that was only slightly higher than cells exposed to wild type toxin. However, the CRAC^Y336P mutant was incapable of killing Jn.9 target cells (87.7 ± 3.3%, p = 0.04). These results indicate that an intact CRAC³³⁶ site is required for LtxA toxicity, whereas the CRAC⁵⁰³ site plays only a limited role in toxicity.

DISCUSSION

LtxA-induced raft clustering is dependent upon a toxin-induced increase in cytosolic Ca²⁺, activation of calpain, and cleavage of talin, which allows mobilization to and subsequent clustering of LtxA and its receptor, LFA-1, in lipid rafts (14). Furthermore, other RTX toxins, such as a leukotoxin (Lkt) produced by M. hemolytica and the adenylate cyclase toxin (CyaA) produced by B. pertussis, have been shown to have similar effects on raft and receptor clustering (16, 47). Here, we describe a unique mechanism of cholesterol binding by LtxA, which we hypothesize to be involved in this previously reported RTX/receptor clustering in lipid rafts.

Cholesterol-rich lipid rafts are dynamic and complex structures that admit or exclude certain proteins; as such, they provide platforms for signaling cascades (48). For example, LFA-1 clustering in lipid rafts plays several important roles in the ontogeny of immune responses, such as the formation of an immunological synapse (49, 50) or the arrest of immune cells on endothelia expressing intercellular adhesion molecules (51). In addition, cholesterol-rich lipid rafts play central roles in the pathogenesis of various microorganisms in many different ways, including docking, internalization, hiding, hijacking cell signaling, and oligomerization (52–59). FimH-expressing E. coli binds to the glycosylphosphatidylinositol-linked protein CD48, which is located in lipid rafts, as a first step in internalization (56). Bacteria, including Campylobacter jejuni and Mycobacterium bovis, as well as viruses are internalized in raft-dependent processes (53–55, 58). A number of bacterial toxins, including streptolysin O, Vibrio cholerae cytolysin, and anthrax toxin, have been shown to use lipid rafts as a concentrating platform for oligomerization (52, 59), whereas thiol-activated/thiol-dependent streptolysins and listerolysin use raft lipids as receptors (57).

The mechanism of lipid raft association and cholesterol binding by LtxA appears to be distinct from these previously reported mechanisms. We have shown that LtxA binds specifically to cholesterol-containing membranes in a unique mechanism with a strong cholesterol-dependent dissociation rate but a cholesterol-independent association rate (Fig. 2). Two regions of the deduced amino acid sequence of LtxA, ³³⁴LEEYSKR³³⁹ (CRAC³³⁶) and ⁵⁰²VDYLKK⁵⁰⁵ (CRAC⁵⁰³), were identified as potential cholesterol binding sites, based upon the CRAC motif, (L/V)X_1–5YX_1–5(R/K) (60, 61). Although both sites are juxtaposed to the LtxA hydrophobic domain, the requirements for a CRAC motif are quite flexible, and therefore, not all predicted CRAC sites bind cholesterol or possess an in vivo function. For this reason, we undertook an experimental approach employing SPR and DSC with LtxA and synthetic peptides corresponding to both CRAC³³⁶ and CRAC⁵⁰³ (Table 5). Our results indicate that LtxA binds cholesterol strongly, and both CRAC peptides CRAC^336WT and CRAC^503WT also interact strongly with cholesterol. However, only CRAC^336WT was able to inhibit binding of LtxA to cholesterol. CRAC^503WT and all of the control peptides had no effect on binding of LtxA to cholesterol. In addition, only CRAC^336WT, the only peptide to inhibit binding to cholesterol, inhibited LtxA-mediated cytotoxicity. This behavior is consistent with our previous work, in which we found that extraction of cholesterol from Jn.9 cells inhibits LtxA-mediated cytotoxicity (14). In the panel of CRAC mutants, modification of CRAC³³⁶ reduced the cytotoxicity of LtxA, whereas modification of CRAC⁵⁰³ had only a limited effect on the cytotoxicity of LtxA.

Together, the results indicate that although both CRAC sites may interact with cholesterol, the first CRAC site, CRAC³³⁶, is responsible for cholesterol binding by LtxA. Interestingly, this first CRAC site is highly conserved among RTX toxins, indicating that this mechanism may be common to other members of this toxin family (Table 4). The second CRAC site, CRAC⁵⁰³, which appears to have a less important role in cholesterol binding and cytotoxicity of LtxA, is less conserved among the RTX toxins, appearing only in LtxA.

Our data indicate that cholesterol binding by LtxA is mediated by CRAC³³⁶; however, we could not rule out the possibility that the acyl groups attached to LtxA could also be involved in the process. For some proteins, acylation is required for membrane binding (62–66); however, the role of acylation in the RTX toxins has not yet been conclusively determined (34). Studies have shown that acylation does not affect the ability of E. coli α-HlyA to bind to either erythrocytes or liposomes (67, 68), but others have shown that acylation is involved in this process (69–71). Alternatively, acylation of the RTX toxins has been proposed to be involved in protein-protein interactions or pore formation (34, 68, 69, 72). Previously, we showed that the acyl chains on LtxA are saturated and/or hydroxylated (10); this hydroxylation may sterically inhibit association with cholesterol-containing membranes, which are more tightly packed than cholesterol-poor membranes. Acylation clearly plays a role in cytotoxicity, because the nonacylated mutant was unable to kill Jn.9 cells (Table 1); however, it appears that the role of acylation in the cytotoxic mechanism is not the binding to cholesterol, because there was no significant difference in the binding of acylated LtxA (LtxA^ltxC/+ltxC) to cholesterol compared with unacylated LtxA (LtxA^ltxC).

LtxA is not the only toxin produced by A. actinomycetemcomitans that binds to cholesterol using CRAC sequences. The cytolethal distending toxin produced by this organism binds membrane cholesterol before gaining access to the cytoplasm, where it triggers cell cycle arrest (43). Surprisingly, although both the A. actinomycetemcomitans-produced cytolethal distending toxin and LtxA bind cholesterol via CRAC sites, the mechanisms of binding of the toxins are quite different. The affinity of the cytolethal distending toxin for cholesterol occurs through an increased association rate when binding to cholesterol and cholesterol-rich rafts (43), suggesting that the cytolethal distending toxin uses cholesterol as a receptor. Conversely, the affinity of LtxA for cholesterol is not dependent upon differences in association rate (k_a); rather, it is influenced by the cholesterol-dependent change in dissociation rate (k_d), suggesting that LtxA is more indiscriminate in its binding, but once bound to cholesterol, the attachment is almost irreversible.

The lipid compositions used in this work were specifically chosen for their raft-forming nature. At room temperature, a membrane composed of DMPC and 0% cholesterol exists in a non-raft-like (liquid-disordered) phase, whereas at 40 and 60% cholesterol, the membrane exists in a raft-like (liquid-ordered phase) (27, 28). In vitro, the lipid composition of lipid rafts is often ∼30% (74–77). The cholesterol compositions used in this work are valid in both an in vitro and in vivo context.

Previous work in our lab found that LtxA is located in lipid rafts only after it has bound to its receptor, LFA-1. Before the binding of LtxA, LFA-1 is located outside of the raft, and upon LtxA binding, the LtxA·LFA-1 complex moves to the raft, causing large scale clustering (14). Here, we demonstrated that the association rate of LtxA to membranes does not change with the cholesterol composition; in other words, LtxA does not have a clear preference for binding to cholesterol, an observation that is consistent with this previous paper, in which LtxA was found in both raft (cholesterol-rich) and non-raft (cholesterol-poor) regions of the membrane. Once LtxA binds to cholesterol, however, it remains tightly bound, providing a possible mechanism for the observed LtxA·LFA-1 clustering in lipid rafts (Fig. 9). Movement of proteins in and out of rafts is often transient (23, 33, 46), and LFA-1 has been shown to reside in two subsets: one within rafts and one outside (73). We hypothesize that LFA-1 is able to move into and out of rafts until it is bound by LtxA, at which point, LtxA binds tightly to cholesterol, holding the LtxA·LFA-1 complex in the raft and allowing the formation of large clusters within the raft. Because we have found that binding of LtxA to cholesterol is essential for LtxA-mediated cell death, we propose that this cholesterol-dependent clustering is a key step in the cytotoxicity of LtxA.

FIGURE 9.

Proposed mechanism of LtxA binding and resulting LFA-1/LtxA clustering in lipid rafts. In the absence of LtxA, LFA-1 moves transiently into and out of lipid rafts (top panel). Binding of cholesterol by LtxA is almost irreversible, and therefore, transient movement of the LtxA·LFA-1 complex into the raft becomes “permanent,” resulting in large LtxA·LFA-1 clusters in the lipid rafts (bottom panel).

The specific binding of other RTX toxins to cholesterol has not been reported previously, although it has been shown that both CyaA and Lkt cause a receptor/toxin clustering in lipid rafts (16, 47). The mechanism of raft clustering by Lkt is similar to what we have reported for LtxA; here, we have shown that Lkt shares CRAC³³⁶ with LtxA, suggesting that the binding of cholesterol by this RTX toxin may drive this receptor-dependent clustering in lipid rafts. However, CyaA, which does not contain either of the two CRAC sites, has been shown to induce the clustering of α_Mβ₂ integrins in lipid rafts (47). The mechanism by which CyaA clusters in rafts differs slightly from the mechanism used by LtxA (47); an interesting question will be whether this difference in mechanism can be explained by the lack of a CRAC site in CyaA or the fact that CyaA binds to a different β2 integrin than does LtxA.

In this work, we have identified a unique but likely conserved mechanism of cholesterol binding by a bacterial toxin. The binding of cholesterol by LtxA occurs in an indirect fashion but appears to regulate the cholesterol dependence of LtxA cytotoxicity.