Role of a disintegrin and metalloprotease 10 inStaphylococcus aureus α-hemolysin–mediatedcellular injuryGeorgia A. Wilkea and Juliane Bubeck Wardenburga,b,1

Departments of aPediatrics and bMicrobiology, University of Chicago, Chicago, IL 60637

Edited by Richard P. Novick, New York University School of Medicine, New York, and approved June 16, 2010 (received for review February 15, 2010)

Staphylococcus aureus α-hemolysin (Hla), a potent cytotoxin, playsan important role in the pathogenesis of staphylococcal diseases,including those caused by methicillin-resistant epidemic strains. Hlais secreted as a water-soluble monomer that undergoes a series ofconformational changes to generate a heptameric, β-barrel struc-ture in host membranes. Structural maturation of Hla depends onits interaction with a previously unknown proteinaceous receptorin the context of the cell membrane. It is reported here that a dis-integrin and metalloprotease 10 (ADAM10) interacts with Hla andis required to initiate the sequence of events whereby the toxin istransformed into a cytolytic pore. Hla binding to the eukaryotic cellrequires ADAM10 expression. Further, ADAM10 is required for Hla-mediated cytotoxicity, most notably when the toxin is present atlow concentrations. These data thus implicate ADAM10 as theprobable high-affinity toxin receptor. Upon Hla binding, ADAM10relocalizes to caveolin 1-enriched lipid rafts that serve as a platformfor the clustering of signaling molecules. It is demonstrated thatthe Hla–ADAM10 complex initiates intracellular signaling eventsthat culminate in the disruption of focal adhesions.

pore-forming cytotoxin | cellular receptor

The Gram-positive extracellular pathogen Staphylococcus au-reus is one of the leading causes of human bacterial infection.

As a commensal of the skin, S. aureus is well positioned to causeinfection of the skin and soft tissues, the bloodstream, and thelower respiratory tract, which are the principal sites of clinicallyrelevant infection (1, 2). To facilitate entry and spread through thehost tissue, S. aureus encodes a number of virulence factors thatallow the organism to breach structural and immunological bar-riers to infection. One of the most prominent and well-charac-terized virulence factors produced by S. aureus is α-hemolysin(Hla), a pore-forming cytotoxin implicated in the pathogenesis ofsepsis, pneumonia, and severe skin infection (3–6). Pore formationon susceptible host cell membranes triggers alterations in iongradients, loss of membrane integrity, activation of stress-signalingpathways, and cell death (3, 7). Hla binds to most eukaryotic cells,often by a nonspecific adsorptivemechanism requiringmicromolarconcentrations of toxin (8). However, a high-affinity interaction ofthe toxin with a proteinaceous eukaryotic receptor has been sug-gested because rabbit erythrocytes are significantly more sensitiveto Hla than human erythrocytes, correlating with the identificationof 1,200–5,000 toxin-binding sites per rabbit cell (8, 9). Binding issaturable and time dependent, consistent with a ligand–receptorinteraction (8, 10). In addition to these data, membrane lipidsseem to be central to the interaction of the toxin with theeukaryotic cell.Membrane cholesterol or sphingomyelin depletionabrogates toxin binding and cytotoxicity, and the addition of ex-ogenous phosphocholine disrupts toxin binding and impairs rabbitred cell hemolysis (11). These data have led to the hypothesis thatclustered phosphocholine head groups serve as the high-affinitybinding site for Hla. Despite intense investigation of this toxin,a proteinaceous cellular receptor has not yet been identified, and itremains unclear how Hla binding exhibits both species specificityand a requirement for particular membrane lipids.

ResultsADAM10 Mediates S. aureus α-Hemolysin Binding to Eukaryotic Cells.We used a biochemical approach to purify a putative Hla re-ceptor, taking advantage of species-specific receptor expression.Rabbit and human erythrocyte ghosts were incubated in the ab-sence of detergent with GST or GST-HlaH35L, a Hla mutant thatprecludes pore formation while preserving membrane binding(12). After toxin treatment, ghosts were solubilized with Triton X-100, and GST or GST-HlaH35L precipitated proteins were ana-lyzed by SDS/PAGE and silver staining. An ≈65-kDa proteinbound GST-HlaH35L from rabbit erythrocytes (Fig. 1A, arrow).Mass spectroscopic analysis of this protein yielded peptides cor-responding to ADAM10 (a disintegrin and metalloprotease 10),a zinc-dependent metalloprotease that is expressed on the surfaceof many distinct cell types as a type I transmembrane protein (Fig.S1, peptides highlighted in bold). GST or GST-HlaH35L boundproteins were then examined by immunoblotting with anti-humanADAM10 (hADAM10) antisera, leading to the detection ofrabbit ADAM10 (rADAM10) specifically in GST-HlaH35L pre-cipitates (Fig. 1B, Upper). Control immunoblots (Fig. 1B, Lower)demonstrating the presence of GST or GST-HlaH35L were pre-pared from samples processed in parallel or probed after removalof the upper section of the nitrocellulose containing ADAM10.Flow cytometric analysis revealed expression of ADAM10 onrabbit erythrocytes and toxin-susceptible human A549 alveolarepithelial cells (Fig. 1C, blue and green solid lines, respectively).Human erythrocytes were devoid of ADAM10 (Fig. 1C, red solidline), in agreement with the species-specific expression patternpredicted for the Hla receptor.To evaluate the Hla–ADAM10 interaction in intact cells, A549

cells were treated with HlaH35L or active Hla. Cells were washedand then lysed in Triton X-100 or SDS/DOC-containing RIPAbuffer, andHla immunoprecipiates were examined for hADAM10.Both forms of Hla precipitated ADAM10 (Fig. 2A), with an in-creased amount of ADAM10 detectable in RIPA lysates as com-pared with Triton X-100 lysates, suggesting localization of thecomplex to detergent-resistant lipid rafts. Hla treatment of A549cells induced the aggregation ofADAM10 in discrete punctae (Fig.2B, Middle Left, arrows). Previous studies have demonstrated anassociation of Hla with caveolin 1, a structural protein of special-ized caveolar lipid rafts. The association of Hla with caveolin 1 isnecessary for toxin oligomerization (13, 14). Upon Hla treatmentof A549 cells, ADAM10-containing punctae colocalize with cav-eolin 1 (Fig. 2B, Middle Right; Manders’ coefficient = 0.15 ± 0.08,

Author contributions: G.A.W. and J.B.W. designed research; G.A.W. and J.B.W. performedresearch; G.A.W. and J.B.W. contributed new reagents/analytic tools; G.A.W. and J.B.W.analyzed data; and J.B.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: jbubeckw@peds.bsd.uchicago.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001815107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1001815107 PNAS | July 27, 2010 | vol. 107 | no. 30 | 13473–13478

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020

PBS; 0.28 ± 0.13, Hla; P < 0.0005). Punctae are not evident whencells are treated with a form of Hla that cannot form stable olig-omers (HlaH35L; Fig. 2B, Bottom). These data together reveal thatHla interacts with ADAM10 in the context of the cell membrane,establishing an association that both parallels the known species-selectivity of the toxin evident in erythrocytes and forms a complexthat seems to localize to caveolin 1-enriched lipid rafts.To further examine the role of ADAM10 as a potential Hla

receptor, we quantified the binding of radiolabeled, active Hla toa panel of human epithelial lines expressing varying amounts ofADAM10. A linear correlation was evident between ADAM10expression and Hla binding (Fig. 2C). In contrast, an inverse cor-relation was derived when toxin binding was compared with ex-pression of the related metalloproteases ADAM9 and ADAM17(Fig. S2A), demonstrating the specificity of binding to ADAM10.We assessed binding of radiolabeledHla toA549 cells treatedwithirrelevant siRNAor two siRNAs targetingADAM10 (ADAM10.1andADAM10.2;Table S1).Hla bound cells treatedwith irrevelantsiRNA (Fig. 2D). This interaction was diminished upon siRNA-mediated knockdown of ADAM10, with residual toxin bindingto knockdown cells likely reflecting incomplete knockdown, be-cause ADAM10 remains detectable on the cell surface aftersiRNAtreatment, as assessed byflow cytometric analysis (Fig. 2E).

Binding was unaltered in ADAM9 and ADAM17 siRNA-treatedcells (Fig. S2 B and C).

ADAM10 Is Required for α-Hemolysin–Mediated Cytotoxicity. Hla-deficient S. aureus strains are impaired in causing cytotoxicity in anA549 coculture model with live staphylococci and are severely at-tenuated in amousemodel ofS. aureuspneumonia (4). To examinethe requirement for ADAM10 in Hla-mediated cytotoxicity, A549cells were treated with irrelevant or ADAM10 siRNAs, theninfected with WT S. aureus Newman, an isogenic Hla− mutant ofS. aureusNewman, or theS. aureusHla− strain complementedwithplasmid-encodedHla (Hla− phla). ADAM10knockdown affordedprotection from S. aureus–induced injury as compared with irrel-evant siRNA-transfected cells (Fig. 3A, WT, Hla− phla). This in-jury was dependent on Hla expression (Fig. 3A, Hla-). ADAM10was also required for cytotoxicity mediated by purified Hla (Fig.3B). The requirement for ADAM10 in Hla-mediated injury wasmost evident at concentrations of toxin in the nanomolar range,corresponding to the addition of 30 or 60 hemolytic units (HU)/mLof toxin (0.375 μM and 0.750 μM, respectively). Incubation of cellswith1.25μMtoxin (120HU/mL) resulted ina smaller, yet significantdifference in cytotoxicity between irrelevant and ADAM10 siRNAtreated cells, whereas a toxin concentrationof 2.5 μM(180HU/mL)led to no observable difference upon ADAM10 knockdown. Thisobservation is highly consistent with previous work that has de-scribed a dual mechanism of toxin binding to Hla-targeted cells—a high-affinity, receptor-dependent binding interaction that prevailsat low toxin concentrations, and a nonspecific binding event thatoccurs at higher toxin concentrations (8). Physiologic lytic injuryoccurring at toxin concentrations of ≈50 HU/mL have been sug-gested to be associated with target cell expression of high-affinitybinding sites (3), in good agreement with the data presented herein.Examinationof toxinoligomerization on the surface ofA549 cells

revealed that ADAM10 is required for the formation of the toxinheptamer, because this product is not detected on themembrane ofcells treated with ADAM10 siRNAs (Fig. 3C). A direct correlationbetween the level of surface expression of ADAM10 and suscepti-bility to intoxication was seen in a panel of human epithelial cells(Fig. 3D), with the relatively toxin-insensitive line 1299 displayinga 3.6-fold reduction in ADAM10 expression compared with A549cells. Expression of ADAM9andADAM17did notmarkedly affectsensitivity to Hla (Fig. S2 D and E).

α-Hemolysin–ADAM10 Complex Alters Integrin-Mediated Cell SignalingEvents.ADAMfamily proteins regulate cellmigration and adhesionthrough the combined actions of the metalloprotease and dis-integrin domains (Fig. S1) (15, 16). The N-terminal enzymatic do-main mediates proteolysis of cell adhesion molecules, disruptinglateral cell junctions to facilitate cell mobility and epithelial barrierremodeling. The adjacent disintegrin domain ofmanyADAMmet-alloproteases harbors a conserved aspartate-rich motif that bindscellular integrins; such ADAM–integrin associations also occur inthe absence of this specificmotif, exerting both pro- and antiintegrinfunction (16). Although ADAM10 lacks this specific motif, a repor-ted association of Hla with β1-integrin (17) raised the interestingpossibility that the Hla–ADAM10 complex may alter integrin-mediated signaling. Integrin heterodimers bind extracellular matrixproteins, coupling signals from the extracellular environment to in-tracellular responses at focal adhesion sites (18). Integrin activationleads to the tyrosine phosphorylation and activation of the FAK andSrc tyrosine kinases (19). FAK and Src in turn phosphorylate theadaptor proteins p130Cas and paxillin, allowing these proteins tofunction as molecular scaffolds that link integrin receptors to Rhofamily GTPases and the actin cytoskeleton (20, 21). This phosphor-ylation-regulated signal transduction cascade permits the dynamicassembly and turnover of focal adhesions. To determine whetherHla alters focal adhesion complex signaling, lysates from A549 cellstransfected with irrelevant or ADAM10 siRNAs were examined for

Fig. 1. Hla interacts with ADAM10 expressed on rabbit red blood cells. (A)Rabbit or human erythrocyte ghost preparations were precipitated with GSTcontrol or GST-HlaH35L and visualized by silver stain analysis. **Purified GSTprotein; *purified GST-HlaH35L fusion. (B) GST or GST-HlaH35L precipitates wereimmunoblotted for human ADAM10 or GST. (C) Flow cytometric analysis ofADAM10 on rabbit and human erythrocytes and human A549 cells.

13474 | www.pnas.org/cgi/doi/10.1073/pnas.1001815107 Wilke and Bubeck Wardenburg

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020

the presence of tyrosine-phosphorylated FAK, Src, p130Cas, andpaxillin (Fig. 4 A–D). Treatment of irrelevant transfectants with Hlaled to the dephosphorylation of all four proteins; this response wasabrogated upon ADAM10 knockdown (Fig. 4 A–D). This activityrequires toxin oligomerization, because theHlaH35Lmutant does notinduce tyrosine dephosphorylation of cellular proteins (Fig. S3).Knockdown of β1-integrin expression by specific siRNAs (Table S1)did not alter Hla binding (Fig. S4A), cellular injury (Fig. S4B), orp130Cas dephosphorylation (Fig. S4C). Further, β1-integrin expres-sion by a panel of epithelial cell lines did not correlate strongly withsusceptibility to the cytotoxic effects of Hla (Fig. S4D), togethersuggesting that Hla binding and cytotoxicity are not directly de-pendent on β1-integrin.

Focal Adhesions Are Disrupted by the α-Hemolysin–ADAM10 Complex.The data just described demonstrate that the Hla–ADAM10complex rapidly alters signaling cascades that contribute to thestability of focal adhesions. In this regard, ADAM10 may functionin a manner analogous to other ADAMmetalloproteases, creatinga functional disruption of integrin signaling. Several bacterial vir-ulence factors target the host cytoskeleton through the direct mod-ification of the phosphorylation state of focal adhesion proteins.The Yersiniae YopH tyrosine phosphatase dephosphorylates FAKand p130Cas (22), uncoupling integrin-mediated signals from in-tracellular signaling machinery and promoting cytoskeletal de-rangements that permit evasion of phagocytosis. The Clostridium

difficile toxin A disrupts focal adhesion signaling through Src in-activation, with the resultant dephosphorylation of FAK and pax-illin leading to colonic epithelial cell detachment (23). To evaluatethe cytoskeletal effects of Hla, siRNA transfected A549 cells weretreated with PBS or Hla, then stained with an antivinculin antibody(red) and phalloidin (green). Actin stress fibers originating fromvinculin-containing focal adhesions are apparent in PBS-treatedcells (Fig. 4E, Left). Hla treatment of irrelevant siRNA transfectedcells induces loss of focal adhesions and actin fibers (Fig. 4E, TopRight), which is prevented by ADAM10 knockdown (Fig. 4E,Middle/Bottom Right). Toxin treatment of irrelevant siRNA treatedcells results in an F/G actin ratio of 1.26, decreased from 1.92 inPBS-treated cells (P= 7.2 × 10−10). The F/G actin ratio after toxintreatment of ADAM10 siRNA treated cells is 1.7, reduced from2.1 (P = 0.004). The absolute change in this ratio in ADAM10siRNA treated cells is blunted, consistent with preservation of theF actin cytoskeleton, as viewed microscopically in Fig. 4E.

DiscussionSeveral features of theHla–ADAM10 interaction identify ADAM10as the likelyproteinaceous cellular receptor for the toxin.First, toxinbinding to eukaryotic cells requires ADAM10 expression. Second,Hla physically interacts with ADAM10 in vivo. Third, ADAM10is required for Hla-mediated cytotoxicity. The requirement forADAM10 is most evident at low toxin concentrations in whicha high-affinity receptor-driven binding event would be essential.

Fig. 2. ADAM10 interactswithHla andmediates toxinbinding tohumanepithelial cells. (A) A549alveolar epithelial cellswere treatedwith nontoxigenicHlaH35L oractive Hla, then Hla immunoprecipitates from cell lysates prepared in Triton X-100 or RIPA lysis buffer were immunoblotted for ADAM10. (B) Hla induces therelocalization of ADAM10 to discrete sites on the cell membrane that colocalize with caveolin 1. Quantification of the number of ADAM10 punctae ≥0.05 μm indiameter averagedon a per-cell basis revealed 0.08± 0.14, PBS; 1.63± 0.48, Hla; 0.13± 0.14, HlaH35L, with error values representing the SD. P = 8.7× 10−6 PBS vs. Hla,P= 1.3× 10−5 Hla vs. HlaH35L. Aminimumof 45 cellswas counted for each treatment condition.Quantification ofADAM10–caveolin 1 colocalizationwas determinedby theManders’ coefficient assessed on 20 distinct images (0.15 ± 0.08, PBS; 0.28 ± 0.13, Hla; P = 0.0005) (Scale bars, 10 μm.) (C) Binding of [35S]-methionine–labeledactive Hla examined as a function of ADAM10 surface expression, quantified as themean fluorescence intensity (MFI) shift by flow cytometric analysis in a panel ofhuman epithelial cells. (D) Binding of [35S]-methionine–labeled active Hla to A549 cells treated with irrelevant or ADAM10-specific siRNAs (designated ADAM10.1and ADAM10.2). *P < 2 × 10−6. (E) Flow cytometric detection of surface expressed ADAM10 in A549 cells.

Wilke and Bubeck Wardenburg PNAS | July 27, 2010 | vol. 107 | no. 30 | 13475

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020

Importantly, an examination of cell lines that naturally expressdiffering amounts of ADAM10 confirms the strong correlationbetween ADAM10 expression and the ability of Hla to bind to themembrane and induce cytotoxic injury. These data shed light on themechanism by which the toxin exhibits species-specificity.Hla pore formation is a complex series of events, initiated by

monomeric toxin binding to the surface of the eukaryotic cell.Multiple lines of evidence confirm the importance of themembranelipid environment in Hla-induced injury, because the membrane-opposed region of the toxin interacts with phosphatidylcholine (24),and cholesterol/sphingomyelin-rich membrane domains are re-quired for cytotoxicity (11). On the basis of these observations,clustered phosphocholine head groups have been hypothesized tobe the toxin receptor (11). Although the requirement for mem-brane lipids in toxin biology has been well described, a modelproposing these as the sole cell-binding moiety fails to explainseveral important observations. Most notably, the species speci-ficity of the toxin implies a proteinaceous receptor, because lip-osomesmirroring rabbit and human red blood cells do not parallelthe sensitivities of these cells (10). Further, protease treatment ofhighly sensitive rabbit red blood cells abrogates toxin binding (10).The identification of ADAM10 as an Hla-interacting protein thatis required for initial toxin binding and multiple cellular cytotoxicinsults provides important insight into how Hla engages the hostcell. Our investigations have revealed that the strong in vivo in-teraction of Hla with ADAM10 must occur in the context of anintact cell membrane, because solubilization of the membranebefore toxin binding precludes the association of these proteins. Itthus seems that Hla demonstrates a dual requirement for bothADAM10 and membrane lipids to establish a high-affinity in-teraction with the eukaryotic cell. This model thereby draws to-gether seemingly disparate observations in the field, solving theparadox of how Hla can simultaneously exhibit cell specificity,which can only be conferred by a proteinaceous receptor, andyet also require membrane cholesterol and sphingomyelin.At present, it has not been possible to dissociate ADAM10

from the membrane environment to demonstrate a direct in vitrointeraction with Hla. Crystallographic studies of a portion of na-tive ADAM10 and the related vascular apoptosis-inducing pro-tein-1 reveal a highly conserved, multidomain C-shaped structurecontaining 17 disulfide bonds and several calcium-binding sites(25, 26). The structural complexity of ADAM10, in concert withthe clear requirement for the membrane context, will likely pre-clude in vitro interaction studies of these proteins until such timethat ADAM10 can be incorporated in its native form into an ar-tificial lipid matrix. It remains possible that ADAM10 requiresa membrane-localized coreceptor to bind Hla. The association ofthe ADAM10–Hla complex in SDS-containing buffer suggestsa direct interaction of these proteins, because multimolecularcomplexes are readily dissociated under these conditions. Simi-larly, we did not observe other proteins to be precipitated fromrabbit erythrocytes with GST-HlaH35L.These data provide a mechanistic view of the assembly of Hla,

suggesting that its initial interaction with ADAM10 and themembrane directs the assembly of the Hla–ADAM10 complex incholesterol/sphingolipid-rich caveolar rafts. This clustering likelyincreases the local concentration of Hla, permitting caveolin1-directed oligomerization of the toxin and providing accessibilityto caveolae-associated proteins FAK and Src that mediate thebiologic effects of Hla (Fig. 5). Focal adhesion disruption by theHla–ADAM10 complex provides a mechanism by which the toxinmay perturb cellular barriers to cause invasive disease and facil-itate superantigen permeation through impenetrable stratifiedcell layers (27). By placing ADAM10 as a proteinaceous Hla re-ceptor, genetic mutation of ADAM10 should confer resistance totoxin-mediated disease. ADAM10 null mice display embryoniclethality (28), thus conditional knockouts will be essential to de-fine the relevance of the Hla–ADAM10 interaction in S. aureus

Fig. 3. ADAM10 is required for Hla-mediated cytotoxicity. (A) Irrelevant orADAM10 siRNA transfected A549 cells were cocultured with WT S. aureus,S. aureus deficient in production of Hla (Hla−), or S. aureusHla− complementedwith plasmid-encoded Hla (Hla− phla), and cell injury measured by lactate de-hydrogenase (LDH) release. *P < 0.008 vs. irrelevant siRNA treated cells. Errorbars delineate the SEM examined in six replicates from two independentexperiments. (B) LDH release from A549 cells treated with the indicated con-centrations of purified, active Hla. *P < 0.0002 vs. irrelevant siRNA treated cells.Error bars delineate the SD of six replicates. (C) Detection of radiolabeled,oligomeric Hla (Hla7) bound to irrelevant or ADAM10 siRNA transfected A549cells. (D) LDH release in a panel of epithelial cell lines cultured with purifiedactive Hla examined as a function of ADAM10 surface expression, quantifiedas the mean fluorescence intensity (MFI) shift by flow cytometric analysis.

13476 | www.pnas.org/cgi/doi/10.1073/pnas.1001815107 Wilke and Bubeck Wardenburg

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020

infection. In addition, inhibitors of zinc metalloproteases, in-cluding ADAM10, are being investigated for clinical potency ina wide array of inflammatory and malignant diseases and mayhave utility as tools to combat S. aureus infection.

MethodsBacterial strains, cell lines and culturing conditions, antibodies, recombinantproteins, buffers, and flow cytometry techniques are detailed in SI Methods.

Cellular Protein Precipitations and Analysis. Rabbit or human red cell ghosts(5 × 109) were incubated 20 min at room temperature with either 2 μg GSTor GST-HlaH35L, whereas intact A549 cells (1 × 107) were incubated with 2 μguntagged HlaH35L or HIS-Hla proteins for 20 min at room temperature forcoprecipitation studies, which are fully described in SI Methods. Mass spec-

trometry was performed by the Taplin Mass Spectrometry Facility at HarvardMedical School.

siRNA Experiments. siRNAs were purchased from Applied BioSciences (TableS1) and used as described previously (29), with minor modifications relevantto A549 cells as detailed in SI Methods.

Radiolabeled Hla Studies. Binding of radiolabeled Hla prepared as describedpreviously (30) was assessed after treatment of 1 × 106 cells/500 μL with 0.3 nMHla in PBS over 5 min. Processing of cells and assessment of toxin oligomeri-zation is detailed in SI Methods.

Immunofluorescence Microscopy. Cells stained for microscopic analysis weretreated for 60 min (ADAM10, caveolin 1 studies) or 2 h (focal adhesionstudies) in F12K media with 30 nM recombinant active Hla or the corre-sponding HlaH35L mutant and stained with the indicated antibodies asdetailed in SI Methods. Microscopy was performed using an Olympus DSUconfocal microscope or a Nikon Eclipse TE2000U microscope. Image pro-cessing and analysis of immunofluorescence microscopy was performedwith ImageJ software (http://rsbweb.nih.gov/ij/). The quantification ofADAM10 punctae was performed with ImageJ software by automatedcounting of particles with a lower size cutoff of 0.05 μM; quantification ofcolocalization as determined by Manders’ coefficient was performed usingJACoP (http://rsb.info.nih.gov/ij/plugins/track/jacop.html). Quantification of theF/G actin ratio was performed on toxin-treated cells according to previouslydescribed methods (31), as detailed in SI Methods.

Statistical Analysis. Statistical significance was calculated using the two-tailedStudent’s t test.

ACKNOWLEDGMENTS. We thank D. Missiakas, O. Schneewind, A. Schwartz,and R. Haselkorn for critical discussions and comments on the manuscript;B. Ragle for optimizing siRNA knockdown in A549 cells and early experimentson radiolabeled Hla binding; G. Randall and N. Heaton for assistance withsiRNA experiments; and J. Kern, V. Bindokas, and C. Labno for microscopy

Fig. 4. ADAM10 is required for Hla-mediated focal adhesion disruption. (A) Cell lysates from irrelevant or ADAM10 siRNA transfected A549 cellstreated with active Hla were probed with an antibody specific for phospho-Tyr397 FAK (Upper) or anti-FAK control (Lower). (B) Lysates prepared asdescribed in A were probed with an antibody specific for phospho-Tyr-416 Src (Upper) or anti-Src control (Lower). (C ) Lysates prepared as described in Awere probed with an antibody specific for phospho-Tyr-165 p130Cas (Upper) or anti-p130Cas control (Lower). (D) Lysates prepared as described in Awere probed with an antibody specific for phospho-Tyr-118 paxillin (Upper) or antipaxillin control (Lower). (E ) Irrelevant (Top) or ADAM10 (Middle,Bottom) siRNA transfected A549 cells were treated with PBS or active Hla and stained to detect vinculin (red) and filamentous actin (phalloidin, green).(Scale bars, 10 μm.)

Fig. 5. Model depicting the ADAM10-Hla complex localized to caveolin 1-enriched membrane domains, causing cytolysis and focal adhesion disrup-tion in toxin-sensitive cells.

Wilke and Bubeck Wardenburg PNAS | July 27, 2010 | vol. 107 | no. 30 | 13477

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020

support. This work was supported by the Deparments of Pediatrics andMicrobiology at theUniversity ofChicago. Theauthors acknowledgemember-

ship in and support from the Region V “Great Lakes” Regional Center ofExcellence (National Institutes of Health Award 2-U54-AI-057153).

1. Lowy FD (1998) Staphylococcus aureus infections. N Engl J Med 339:520–532.2. Klevens RM, et al.; Active Bacterial Core surveillance (ABCs) MRSA Investigators (2007)

Invasive methicillin-resistant Staphylococcus aureus infections in the United States.JAMA 298:1763–1771.

3. Bhakdi S, Tranum-Jensen J (1991) Alpha-toxin of Staphylococcus aureus.Microbiol Rev55:733–751.

4. Bubeck Wardenburg J, Schneewind O (2008) Vaccine protection against Staphylococcusaureus pneumonia. J Exp Med 205:287–294.

5. Menzies BE, Kernodle DS (1996) Passive immunization with antiserum to a nontoxicalpha-toxin mutant from Staphylococcus aureus is protective in a murine model.Infect Immun 64:1839–1841.

6. Kennedy AD, et al. (2010) Targeting of alpha-hemolysin by active or passiveimmunization decreases severity of USA300 skin infections in a mouse model. J InfectDis, in press.

7. HusmannM, et al. (2006) Differential role of p38 mitogen activated protein kinase forcellular recovery from attack by pore-forming S. aureus alpha-toxin or streptolysin O.Biochem Biophys Res Commun 344:1128–1134.

8. Hildebrand A, Pohl M, Bhakdi S (1991) Staphylococcus aureus alpha-toxin. Dualmechanism of binding to target cells. J Biol Chem 266:17195–17200.

9. Cassidy PS, Harshman S (1973) The binding of staphylococcal 125I-alpha-toxin (B) toerythrocytes. J Biol Chem 248:5545–5546.

10. Cassidy P, Harshman S (1976) Studies on the binding of staphylococcal 125I-labeledalpha-toxin to rabbit erythrocytes. Biochemistry 15:2348–2355.

11. Valeva A, et al. (2006) Evidence that clustered phosphocholine head groups serve assites for binding and assembly of an oligomeric protein pore. J Biol Chem 281:26014–26021.

12. Menzies BE, Kernodle DS (1994) Site-directed mutagenesis of the alpha-toxin gene ofStaphylococcus aureus: Role of histidines in toxin activity in vitro and in a murinemodel. Infect Immun 62:1843–1847.

13. Vijayvargia R, Suresh CG, Krishnasastry MV (2004) Functional form of Caveolin-1 isnecessary for the assembly of alpha-hemolysin. Biochem Biophys Res Commun 324:1130–1136.

14. Pany S, Vijayvargia R, Krishnasastry MV (2004) Caveolin-1 binding motif of alpha-hemolysin: Its role in stability and pore formation. Biochem Biophys Res Commun 322:29–36.

15. Reiss K, Saftig P (2009) The “a disintegrin and metalloprotease” (ADAM) family ofsheddases: Physiological and cellular functions. Semin Cell Dev Biol 20:126–137.

16. White JM (2003) ADAMs: Modulators of cell-cell and cell-matrix interactions. CurrOpin Cell Biol 15:598–606.

17. Liang X, Ji Y (2007) Involvement of alpha5beta1-integrin and TNF-alpha inStaphylococcus aureus alpha-toxin-induced death of epithelial cells. Cell Microbiol 9:1809–1821.

18. Streuli CH, Akhtar N (2009) Signal co-operation between integrins and other receptorsystems. Biochem J 418:491–506.

19. Cary LA, Guan JL (1999) Focal adhesion kinase in integrin-mediated signaling. FrontBiosci 4:D102–D113.

20. Nojima Y, et al. (1995) Integrin-mediated cell adhesion promotes tyrosinephosphorylation of p130Cas, a Src homology 3-containing molecule having multipleSrc homology 2-binding motifs. J Biol Chem 270:15398–15402.

21. Vuori K, Hirai H, Aizawa S, Ruoslahti E (1996) Introduction of p130cas signalingcomplex formation upon integrin-mediated cell adhesion: A role for Src familykinases. Mol Cell Biol 16:2606–2613.

22. Persson C, Carballeira N, Wolf-Watz H, Fällman M (1997) The PTPase YopH inhibitsuptake of Yersinia, tyrosine phosphorylation of p130Cas and FAK, and the associatedaccumulation of these proteins in peripheral focal adhesions. EMBO J 16:2307–2318.

23. Kim H, Rhee SH, Pothoulakis C, LaMont JT (2009) Clostridium difficile toxin A bindscolonocyte Src causing dephosphorylation of focal adhesion kinase and paxillin. ExpCell Res 315:3336–3344.

24. Song L, et al. (1996) Structure of staphylococcal alpha-hemolysin, a heptamerictransmembrane pore. Science 274:1859–1866.

25. Janes PW, et al. (2005) Adammeets Eph: An ADAM substrate recognition module actsas a molecular switch for ephrin cleavage in trans. Cell 123:291–304.

26. Takeda S, Igarashi T, Mori H, Araki S (2006) Crystal structures of VAP1 reveal ADAMs’MDC domain architecture and its unique C-shaped scaffold. EMBO J 25:2388–2396.

27. Brosnahan AJ, Mantz MJ, Squier CA, Peterson ML, Schlievert PM (2009) Cytolysinsaugment superantigen penetration of stratified mucosa. J Immunol 182:2364–2373.

28. Hartmann D, et al. (2002) The disintegrin/metalloprotease ADAM 10 is essential forNotch signalling but not for alpha-secretase activity in fibroblasts. HumMol Genet 11:2615–2624.

29. Berger KL, et al. (2009) Roles for endocytic trafficking and phosphatidylinositol4-kinase III alpha in hepatitis C virus replication. Proc Natl Acad Sci USA 106:7577–7582.

30. Ragle BE, Bubeck Wardenburg J (2009) Anti-alpha-hemolysin monoclonal antibodiesmediate protection against Staphylococcus aureus pneumonia. Infect Immun 77:2712–2718.

31. GogartenW, Emala CW, Lindeman KS, Hirshman CA (2001) Oxytocin and lysophosphatidicacid induce stress fiber formation in human myometrial cells via a pathway involving Rho-kinase. Biol Reprod 65:401–406.

13478 | www.pnas.org/cgi/doi/10.1073/pnas.1001815107 Wilke and Bubeck Wardenburg

Dow

nloa

ded

by g

uest

on

Mar

ch 1

5, 2

020