RTX Toxins Recognize a β2 Integrin on the Surface of Human Target Cells*

Actinobacillus actinomycetemcomitans leukotoxin and Escherichia coliα-hemolysin are RTX toxins that kill human immune cells. We have obtained a monoclonal antibody (295) to a cell surface molecule present on toxin-sensitive HL60 cells that can inhibit cytolysis by both RTX toxins. Utilization of this monoclonal antibody for immunoaffinity purification of detergent-solubilized target cell membranes yielded two polypeptide chains of approximate molecular masses of 100 and 170 kDa. Microsequencing of tryptic peptides from the two proteins showed complete homology with CD11a and CD18, the two subunits of the β2 integrin, lymphocyte function-associated antigen 1 (LFA-1). Anti-CD11a and CD18 monoclonal antibodies also inhibited RTX toxin-mediated cytolysis. Direct binding experiments demonstrated the ability of an immobilized RTX to bind LFA-1 heterodimers present in a detergent lysate of human HL60 target cells. Transfection of CD11a and CD18 integrin genes into a cell line (K562) that is not sensitive to either RTX toxin resulted in LFA-1 expressing cells, KL/4, that were sensitive to both toxins. These experiments identify LFA-1 as a cell surface receptor that mediates toxicity of members of this family of pore-forming toxins.

A paradigm emerging from studies of bacterial protein toxins is that target cell recognition is often the initial event of a multistep process ultimately leading to diverse mechanisms of cell death. For example, colicins (1) bind to various outer membrane proteins of Escherichia coli (2), Bacillus thuringiensis ␦-endotoxin recognizes the brush border of insect gut cells (3), diphtheria toxin binds to a heparin-binding epidermal growth factor-like precursor (4), and the family of AB 5 toxins (e.g. cholera toxin) bind to cell surface ganglioside lipids (5). Actinobacillus actinomycetemcomitans leukotoxin is a member of the RTX toxin family of pore-forming hemolysins/leukotoxins (6 -11). Leukotoxin kills cells of the lymphocytic and monomyelocytic lineage of man and some higher nonhuman primates. The narrow host range of leukotoxin suggests that the toxin may bind a specific cell surface receptor. However, no target cell receptors for RTX toxins have yet to be identified. Both toxin-sensitive and several toxin-resistant cells (12) are capa-ble of absorbing toxin from culture supernatants, indicating that simple adsorption of the toxin onto a cell membrane is not enough to cause cell death. The ability of these toxins to adhere to various cell surfaces also indicates that classical experiments designed to detect the presence of a receptor such as saturation of binding or inhibition of labeled toxin binding with excess unlabeled toxin will not be effective in detecting cell surface receptors for this family of toxins. We have employed an alternative approach to identify a cell surface molecule that mediates RTX toxin cytotoxicity.

MATERIALS AND METHODS
Production of Monoclonal Antibodies to HL60 Cell Surface Molecules-Balb/cJ female mice (Jackson Laboratories, Bar Harbor, ME) 12-16 weeks old were immunized intravenously with 2 ϫ 10 7 HL60 cells, boosted on days 10, 20, and 30 with an equal number of cells and allowed to rest for 60 days. 3 days before fusion, the animals again received 2 ϫ 10 7 HL60 cells intravenously. On the day of fusion, spleens were removed aseptically, and a single cell suspension was prepared with a loose fitting tissue homogenizer. The cells were washed once in Dulbecco's minimal essential medium, and the erythrocytes were lysed with 0.17 M ammonium chloride-Tris buffer. Spleen cells recovered in this manner had a viability of Ͼ95% as assessed by trypan blue exclusion. Sp2/0-Agl4 myeloma cells were mixed with spleen cells (1:10) and centrifuged. A 1-ml portion of 30% polyethylene glycol solution (PEG 1000; J. T. Baker Chemical Co., Phillipsburg, NJ) was slowly added to the cell pellet. The pellet was gently stirred, allowed to set for 1 min, and then dispersed by the addition of 50 ml of Dulbecco's minimal essential medium. After centrifugation, the cells were suspended in 30 ml of Kennett's HY medium (4.5 g/liter Dulbecco's minimal essential medium with high glucose, glutamine, 10% NCTC 109, 20% fetal bovine serum, 1% 0.15 mg/ml oxaloacetate, 0.05 mg/ml pyruvate, and 0.2 unit/ml bovine insulin) and 1% 5 mM hypoxanthine and 0.8 mM thymidine), and 0.1-ml portions were placed in 96-well tissue culture plates. The next day, an additional 0.1 ml of medium containing aminopterin (0.04 M) was added to each well. Cells were fed every 3-4 days by drawing off 0.1 ml of spent medium. Clones were visible 7-9 days after the fusion, and when they covered approximately one-half of the bottom of the well, they were screened for their ability to inhibit RTX-mediated cytotoxicitv in the biological assay (13,14).
Analysis of RTX-mediated Target Cell Cytotoxicity-Leukotoxic ac-tivity of recombinant ltxA and hylA gene products was determined essentially as described previously (13). Briefly, E. coli containing ltxC and either ltxA or hylA in the pOTSNco12 plasmid were grown to late log phase, induced, and sonicated. The bacterial sonicate was incubated with target cells at 37°C for 45 min. RTX-sensitive HL60 cells were cultured in RPMI 1640 and used as positive controls, whereas negative controls consisted of incubating target cells with: (i) tissue culture medium, (ii) sonicates from uninduced bacteria, or (iii) sonicates from induced bacteria that contained the pOTSNco12 plasmid without an insert. The cells were placed on ice, 100 l of trypan blue (0.4%) were added, and surviving cells were counted in a hemocytometer. At least four fields were counted in triplicate and averaged for each dilution assayed. The percentage of lysis was calculated by dividing the number of surviving cells by the number of cells in the negative controls and subtracting from 100. Staining and Analysis of Cells for Flow Cytometry-Prior to incubation with anti-integrin antibodies, HL60, KL/4, and K562 cells were washed twice in cold 1% BSA PBS and incubated with mouse IgG myeloma protein (10 g; 30 min on ice) that was the same isotype as the anti-integrin antibody. Immunoglobulin bound during this incubation occurs via the F(c) receptor. After washing again in cold 1% BSA PBS, the cells were incubated with an anti-integrin antibody for 15 min, washed, and incubated for an additional 15 min with FITC-labeled goat anti-mouse IgG (Boehringer Mannheim Biochemical). The fluorescence level of a control group composed of cells, mouse IgG myeloma protein, and FITC-labeled goat anti-mouse IgG represented the background staining due to Fc-bound Ig. The fluorescence level of our experimental groups were then gated on the control group to determine the amount of anti-integrin antibody bound on the cells. After incubations, cells were washed, fixed in 2% paraformaldehyde, and stored at 4°C until analysis. 5000 -10000 cells were read per sample on a FACStar Plus (ABI, Mountain View CA) flow cytometer.
The proteins were run on 8% SDS-PAGE gels using Bio-Rad reagents. Gel solutions were filtered (0.2 m) and stored at 4°C prior to use. Gels (including the stacker) were cast and allowed to stand for 24 -48 h at room temperature (22°C) prior to use, and 0.1 mM thioglycolate was added to the upper chamber buffer prior to electrophoresis. Eluted proteins were solubilized in 5 ϫ buffer containing 0.5 M sucrose and heated at 37°C for 15 min. After electrophoresis, proteins were transferred to a polyvinylidene difluoride membrane. Prior to transfer, polyvinylidene difluoride membrane was pre-wet in MeOH for 10 s and then in transfer buffer with 10% MeOH for 5 min. After transfer, the membrane was rinsed with Milli-Q® water three times for 5 min. The membrane was stained with Amido Black, and proteins of 100 and 170 kDa were cut from the membrane and digested with trypsin. The resultant tryptic fragments were separated by HPLC, and peptides from each protein were selected for sequencing.
Direct Binding of RTX Toxin and ␤2 Integrin Heterodimers-Purified A. actinomycetemcomitans leukotoxin (20) was dissolved in PBS (20 g/ ml) and incubated with 1 ⁄8-inch polystyrene beads (10 beads/20 g of toxin) in a 3-ml siliconized Vacutainer tube (no additives) with gentle rocking overnight at 4°C. The beads were washed with PBS and incubated with 1 mg/ml BSA in PBS for 1 h to block the remaining sites. CHAPS extracts (10 g of protein) of HL-60 cells were incubated with the derivatized beads for 15 h at 4°C in PBS. After incubation, the beads were washed three times with PBST (PBS containing 0.1% Tween 20) followed by a final rinse in 10 mM Tris-HCl, pH 7.4. Bound protein was eluted from the beads using SDS gel loading buffer and size fractionated on 8 -16% SDS gradient gels under nonreducing conditions. Proteins were transferred to a nitrocellulose membrane, and the membrane was blocked in PBS, 5% milk and subsequently incubated with anti-LFA-1 antibodies for 15 h at 4°C. The membranes were washed three times with PBST and incubated with a goat anti-mouse HRP conjugate in PBS, 5% milk for 1 h at room temperature. The blots were washed with PBST and developed using chemiluminescent detection.
Transfection of K562 with ␤2 Integrin Genes-K562 transfectants (21) were produced by subcloning the CD11a and CD18 genes into a derivative of the expression vector EE6hCMV (22) carrying a G418 resistance marker.
Reverse Transcriptase Polymerase Chain Reaction-After transfection, the transcription of the integrin genes (23,24) was confirmed by the reverse transcriptase-PCR using primers synthesized in our laboratory. Glyceraldehyde-3-phosphate dehydrogenase (GADPH), which is present in both normal and transfected cells, was used as a control gene (25). Messenger RNA was prepared from KL4 and K562 cell lines using the Quick Prep micro mRNA purification kit (Pharmacia 27-9255-01) followed by first strand cDNA synthesis using You-Prime First-Strand Beads (Pharmacia 27-9264-01) according to the manufacturer's instructions. Subsequent amplification was performed using a Perkin-Elmer 9600 Thermocycler and Ready To Go PCR beads (Pharmacia 27-9555-01) under the following conditions. The primers for PCR were: forward primer CD11a, GGGAATGACCTTGGCAACAGACCCCACAGAT; reverse primer CD11a, GGGTCTCCTGACTCTCCTTGGTCT; forward primer CD18, ATCCTGACTCCATTCGCTGC; reverse primer CD18, CTCGGTCTGAAACTGGTTGG; forward primer GADPH, CCACCCAT-GGCAAATTCCATGGCA; reverse primer GADPH, TCTAGACGGC-AGGTCAGGTCCACC.
The samples were heated to 94°C for 6 min 15 s followed by 30 cycles at 94°C for 15 s, 55°C for 15 s, and 72°C for 15 s and finally an extension of 9 min 15 s at 72°C. Southern blotting was performed using Gene Screen Plus hybridization transfer membrane (NEN Life Science Products NEF-986) according to the manufacturer's instructions. Internal oligomers (23,24) were labeled using the ECL 3Ј-oligolabeling and detection system (Amersham Corp. RPN 2131). The internal probes for Southern blotting were: CD11a, CGCGACAGAGGTGTTCCGGGAGG-AGC (expected PCR product ϭ 610 bp); CD18, GGACCTCTCCTACTC-CATGCTTGATGACCTCAGG (expected PCR product ϭ 496 bp).

Identification and Analysis of a Monoclonal Antibody That
Inhibits RTX-mediated Cytotoxicity-Initial experiments were designed to identify a monoclonal antibody that has specificity for a cell surface molecule on target cells and is capable of inhibiting RTX-mediated cytotoxicity. Balb/cJ mice were immunized with HL60 cells via the intravenous route. Spleens from immunized mice were removed and fused with Sp2/0-Ag14, a mouse myeloma cell line. Hybridoma culture supernatants were screened for their ability to inhibit HL60 cytotoxicity by two RTX toxins, E. coli ␣-hemolysin, and A. actinomycetemcomitans leukotoxin. Of 2,500 clones screened, one clone, mAb295 (IgG1,), was noted to be an efficient inhibitor of the cytotoxic activity of both toxins (Fig. 1A). Fluorescent cell sorting analysis demonstrated that the mAb295 antigen resides on the cell surface of human peripheral blood polymorphonuclear leukocytes, monocytes, and lymphocytes (Fig. 1B).
Cell Surface Molecule Recognized by an RTX-inhibiting Monoclonal Antibody Is a ␤2 Integrin-To identify the mAb295 antigen, HL60 cell membrane extracts were first solubilized with 0.5% CHAPS, passed through a CNBr Sepharose 4B-mAb295 affinity column (26), and then analyzed with SDS-PAGE. Examination of the gels revealed two proteins of relative molecular masses of 100 and 170 kDa had bound to the mAb affinity resin. (Fig. 2, lane 3). No evidence of these bands was noted when the same HL60 CHAPS lysate was passed over an affinity column with a control myeloma protein attached (Fig. 2, lane 4).
Eluted proteins from a mAb295 affinity column were subjected to preparative SDS-PAGE and blotted onto polyvinylidene difluoride membranes. After staining with Amido Black, proteins of interest were cut from the blot and subjected to trypsin digestion. Peptides were separated by HPLC and sequenced by solid phase amino acid sequencing. Two tryptic peptides from the 170-kDa protein were sequenced and showed complete homology with sequences from CD11a, whereas three peptides from the 100-kDa band were sequenced and found to be identical to sequences of CD18 (Table I). The CD11a and CD18 polypeptides form the ␣ and ␤ subunits, respectively, of a leukocyte-specific ␤ 2 integrin, LFA-1.
Anti-CD11a and anti-CD18 Monoclonal Antibodies Are Capable of Inhibiting RTX-mediated Cytotoxicity-Identification of LFA-1 as the molecule that is recognized by mAb295 led us to explore the possibility that anti-CD11 and anti-CD18 monoclonal antibodies might also inhibit RTX-mediated cytotoxicity. A panel of anti-LFA-1 monoclonal antibodies was screened to identify those inhibiting RTX toxin-mediated cytotoxicity (Fig.  3A). TS1/22, an anti-CD11a antibody, and three anti-CD18 antibodies (TS1/18, KIM127, and KIM185) were capable of inhibiting RTX cytolysis. The possibility that the experimental observations were due to Fc receptor binding was ruled out by incubating cells with an isotype-matched (IgG1,) control antibody, MOPC 21. If inhibition of RTX cytotoxicity was due to anti-integrin antibodies binding via the Fc receptor and not to the ␤2 integrin, then MOPC21 would be equally effective in-hibiting target cell killing as the anti-integrin antibodies. This was not observed.
RTX Toxins Bind to LFA-1-Several mechanisms could account for inhibition of RTX-mediated cytotoxicity by anti-␤2 integrin monoclonal antibodies. A likely explanation is steric hinderance. The monoclonal antibody then interferes with binding of the toxin to the same integrin heterodimer. Alternatively, the toxin does not bind to ␤2 integrins, but rather the interaction of antibody and integrin effect changes in the cell surface, rendering it impervious to toxin binding at a distal site. To resolve this issue, a series of experiments examining the ability of RTX toxins to bind directly to ␤2 integrins was initiated. Purified A. actinomycetemcomitans leukotoxin was passively adsorbed onto polystyrene beads, and additional protein binding was blocked by incubating the beads with BSA. The toxin beads were then incubated with a 0.5% CHAPS lysate from HL60 cells. Bound protein was eluted from the beads, run on SDS-PAGE under nonreducing conditions, blotted onto nitrocellulose, and incubated with either 25.3.1 (an anti-CD11a antibody) or KIM185 (an anti-CD18 antibody). The Western blot shows that CD11a (Fig. 4A, lane 3) and CD18 (Fig. 4A, lane 7) polypeptides are detected in the eluate from RTX toxin beads. Control beads coated with BSA alone showed no evidence of binding of either integrin unit (Fig. 4A, lanes 2  and 6). Furthermore, when 10 g of soluble A. actinomycetemcomitans leukotoxin was added to the CHAPS lysate, the ability of the RTX toxin beads to bind either component of the LFA-1 heterodimer was completely abolished (Fig. 4A, lanes 4  and 8). To further assess binding specificity, we determined the ability of mAb295 to block LFA-1 toxin interaction (Fig. 4B). Preincubation of HL-60 lysates with mAb295 (10 g/tube) diminished integrin binding to beads (Fig. 4B, lanes 4 and 8). No inhibition of binding was observed when the lysates were incubated with an equivalent amount of an isotype control antibody, MOPC21 (Fig. 4B, lanes 3 and 7).
Characterization of Transfected K562 Cells-A transfected cell line was employed to further define the involvement of CD18 and CD11a genes in RTX toxin-mediated cytotoxicity. K562, a human erythroleukemia cell line that does not express detectable CD11a or CD18 was transfected with both CD11a and CD18 in a single plasmid. One of the transfected cell lines, KL/4, was shown to express both CD11a and CD18 genes using reverse transcriptase-PCR. Analysis of cDNA revealed the expected 610-and 496-bp PCR products for CD11a (Fig. 5, lane 1) and CD18 (Fig. 5, lane 2) genes in KL/4 cells. No signal for CD11a or CD18 was detected in mock transfected control K562/ gen cells (Fig. 5, lanes 4 and 5). Both cell lines produced detectable levels of the control gene, GADPH (Fig. 5, lanes 3 and

TABLE I Amino acid sequences of tryptic peptides derived from 100-and 170-kDa proteins present in the eluate from a mAb295-Sepharose column
Eluates from the mAb295 immunoaffinity column were subjected to SDS-PAGE and transferred to polyvinylidene difluoride membranes. Protein bands (100 and 170 kDa) were cut from the blots and digested with trypsin, and peptides isolated via HPLC for sequencing. Sequences were then used to query the protein data base, and homologous sequences were identified.

6).
The 610-and 496-bp PCR products amplified from KL/4 hybridized to internal sequence probes of CD11a (Fig. 5, lane a) and CD18 (Fig. 5, lane b), respectively. FACS analysis of transfected KL/4 cells (Table II) demonstrated cell surface expression of CD11a and CD18 gene products, whereas no evidence of LFA-1 expression was observed on K562 cells. Fluorescence intensity levels of the two proteins appeared similar, indicating that production of the two pro-teins is comparable. In addition, the failure to demonstrate expression of either CD11b or CD11c chains demonstrate that LFA-1 is the only ␤2-integrin expressed on KL/4 cells. Functional characterization of the LFA-1 on cells (KL/4) has been described previously (21). Integrin expression by KL/4 cells appears to be in a low state of activation as determined by their inability to bind either ICAM-1 ligand or antibodies that recognize activation epitopes.
Influence of ␤2 Integrin Gene Expression on the Susceptibility of Cells to RTX Toxins-K562 cells do not express ␤2 integrin gene products and are resistant to the lytic effects of E. coli ␣-hemolysin or A. actinomycetemcomitans leukotoxin (12). Further, K562/gen, a mock transfected K562 cell control, showed no evidence of susceptibility to either toxin (Fig. 6A). However, when K562 cells were transfected with CD11a and CD18 genes on the same plasmid, the resultant cell line (KL/4) was sensitive to both RTX toxins (Fig. 6A). The lethal effects of the RTX toxins could be ablated by preincubation of KL/4 cells with anti-CD11a monoclonal antibody mAb295 but not by MOPC21, an isotype matched control antibody.

TABLE II
Expression of ␤2 integrin heterodimers in transfected K562 cells FACS analysis of K562 cells and LFA-1 (KL/4)-transfected cells were stained with anti-␤2 integrin antibodies. Antibodies were used at a concentration of 10 g/ml, and staining was done at 4°C. Results are shown as mean fluorescence. ments. K562 cells incubated with A. actinomycetemcomitans leukotoxin remain uniform in size and shape with prominent cellular organelles and are morphologically similar to untreated cells (Fig. 6B, plate 1). In contrast, KL/4 cells showed significant morphologic changes after toxin exposure. In half of the cells we noted an increase in size associated with vacuolated cytoplasm and degenerating organelles (Fig. 6B, plate 2, arrow). The remainder of KL/4 cells showed no evidence of RTX intoxication. KL/4 cells preincubated with mAb295 (10 g/ml) were morphologically similar to untreated controls (data not shown). DISCUSSION We have used mAbs, direct binding, and gene transfection experiments to demonstrate that A. actinomycetemcomitans leukotoxin and E. coli ␣-hemolysin bind to CD11a-CD18 integrin (LFA-1) on human target cells. LFA-1 is found on most circulating leukocytes (lymphocytes, neutrophils, monocytes, and macrophages) but not on cells of nonhematopoetic origin. A member of the ␤2 integrin family, LFA-1, mediates leukocyte binding to endothelial cells via interactions with several members of the ICAM family (28), a function critical to leukocyte emigration during inflammatory responses. In addition, LFA-1 is involved in T cell functions such as adhesion of cytotoxic T cells to their target cells, antigen-induced T cell proliferation, and interactions between B cells and T cells.
The eukaryotic host cell membrane contains many molecules that could provide an adhesive function for invading pathogens, and several members of the integrin family have been shown to serve as targets for a variety of virulence determinants (29). CD11b-CD18 (MAC-1), a ␤2 integrin that binds the C3bi complement component, is used by several pathogens. Microorganisms that bind to MAC-1 are taken into the immune cell by the integrin and bypass powerful killing mechanisms, such as H 2 O 2 or oxygen free radicals, associated with internalization by the phagocytic pathway (30). Unlike other ␤2 integrins, LFA-1 does not undergo endocytosis (31). The current study demonstrates a unique interaction whereby RTX toxins in which a ␤2 integrin is utilized destroy host immune cells rather than facilitate microbial internalization.
Our results demonstrating direct binding of A. actinomyce-temcomitans leukotoxin to the LFA-1 heterodimer (Fig. 4) supports the mAb studies (Fig. 3) in identifying this integrin as a receptor for RTX toxins. Toxin-coated polystyrene beads bound both CD11a and CD18 subunits when analyzed by Western blotting. Furthermore, the fluorographs provided several interesting insights into the toxin-integrin interaction. The anti-CD18 antibody detected a primary band at 100 kDa and a smaller band at 90 kDa. The presence of the two ␤2 integrin species has been described previously (32,33) and represents the mature polypeptide and precursor of the ␤ chain. It appears that the leukotoxin is able to bind to LFA-1 molecules containing the precursor form of CD18. Examination of the fluorograph produced with anti-CD11a antibody showed relatively small amounts of ␣ L chain present in the eluate from the toxin beads when compared with staining of the input protein. Additional experiments are required to determine if this observation represents preferential binding of CD18 molecules to the beads, inefficient elution of CD11a from the toxin beads, or partial denaturation of CD11a during binding, elution, or electrophoresis such that antibody reactivity is diminished. The identification of an integrin receptor as the receptor for RTX toxins necessitates consideration of the role that molecular conformation may be playing in the toxin-target cell interaction. A feature of many integrins is their capacity to undergo stimulus-induced conformational changes that lead to increased ligand binding. Although the mechanism for such conformational changes is not currently understood, a variety of stimuli can produce increased integrin/ligand interactions (34) through modulation of the conformational state of the integrin (16). The affinity changes associated with LFA-1 activation appear to be complex and involve at least three conformational states (35). Thus, it is possible that different activation states (36,37) could alter sensitivity to RTX toxins. The current studies suggest that the activation state of the integrin may influence the integrin-toxin interaction, because LFA-1 expressed on KL/4 cells is in a low activation state (21). KL/4 cells are less sensitive than HL60 to both RTX toxins when compared with HL60 cells under the same conditions (Fig. 6A). Supporting the low state of reactivity are the observations that the cells do not bind either ICAM-1 or antibodies that have been shown to recognize activation epitopes. We are currently determining if the differences in toxin sensitivity we observed are due to an unfavorable integrin binding state in KL/4 cells or due to differences unrelated to LFA-1 expression.
Regardless of the effect that integrin activation states may have on toxin-target cell interactions, our results demonstrate that RTX toxins do not bind in a manner homologous to the LFA-1-ICAM-1 interaction. KIM 127 (Fig. 3A) inhibited the target cell killing of both RTX toxins. Previous studies (21) have shown that binding of KIM 127 to cells activates LFA-1 function by increasing receptor binding to its ligand, ICAM-1. If RTX toxins did bind to LFA-1 at the canonical ICAM-1 binding site, we would expect KIM127-treated cells to have increased sensitivity to both toxins. Furthermore, failure of soluble ICAM-1 to inhibit RTX cytolysis, indicates that RTX toxins recognize regions of LFA-1 not involved in ligand binding. 2 RTX toxins are subdivided into two classes based upon their target cell specificity. A. actinomycetemcomitans leukotoxin kills a narrow range of target cell types including lymphoid cells and granulocytes from man and certain primates (38,39). The distribution pattern of CD11a-CD18 coincides with the range of cells susceptible to A. actinomycetemcomitans leukotoxin. In contrast, E. coli ␣-hemolysin is an RTX hemolysin that is toxic to a wide range of cell types such as red blood cells found in many species, chicken embryo fibroblasts, rabbit granulocytes, and mouse fibroblasts (3T3 cells) (40). In the present experiments, the finding that anti-CD11a and CD18 monoclonal antibodies inhibited both toxins suggests that expression of LFA-1 was sufficient for killing. However, the capacity of E. coli ␣-hemolysin to kill a wider range of cells may reside in its ability to bind cell surface molecules other than LFA-1. We are currently exploring this possibility.
The function that cell surface receptor molecules play in the pathogenesis of other bacterial toxins provides important paradigms in attempting to understand the RTX toxin receptor (1,(3)(4)(5). Recognition of cell surface molecules by RTX toxins represents the crucial component of a complex multi-step process in which the toxin transits from an aqueous to an amphipathic form. Bacterial protein toxins must be soluble in aqueous solutions while maintaining a capacity to interact within the hydrophobic environment of biological membranes. To achieve this dichotomous existence, toxins are organized into domains, and these regions are folded in such a way that the region of the toxin ultimately interacting with the cell membrane is internalized when the protein is in its aqueous form (41). Although there is a natural energy barrier that resists unfolding, toxin binding to LFA-1 could provide a very efficient mechanism to lower this barrier and catalyze a partial unfolding of the molecule leading to exposure of sequestered hydrophobic residues resulting in a membrane-inserted pore structure.
In conclusion, our results show that a member of the ␤2integrin family, LFA-1, is a cell surface receptor for RTX toxins on human target cells. The ability of ␤2 integrin monoclonal antibodies to inhibit cell killing by RTX toxins and of ␤2 integrin gene expression to convert a toxin-insensitive cell to a toxinsensitive one provide strong support for hypotheses linking ␤2 integrins to RTX-mediated cytotoxicity. Diapedesis toward sites of infection caused by either A. actinomycetemcomitans or uropathogenic E. coli requires participation of the CD11-CD18 adhesion complex. The ability of these pathogens to utilize these critical cell adhesion molecules to destroy host immune cells at mucosal sites would thus allow microbial colonization to proceed unabated. Continued study of the mechanisms involved in RTX toxin cytotoxicity will facilitate a further understanding of the biology of this interesting group of toxins and could lead to improved diagnostic and therapeutic applications.