Heterodimeric integrin complexes containingβ1-integrin promote internalization andlethality of anthrax toxinMikhail Martchenkoa, Sun-Young Jeonga, and Stanley N. Cohena,b,1

Departments of aGenetics and bMedicine, Stanford University School of Medicine, Stanford, CA 94305

Contributed by Stanley N. Cohen, July 22, 2010 (sent for review January 15, 2010)

To kill macrophages, the lethal factor component of Bacillus an-thracis toxin binds to a carrier protein (PA), which then interactswith the CMG2 receptor protein on the cell surface and is endocy-tosed into the cytoplasm. CMG2, as well as TEM8, a second PA re-ceptor not present on macrophages, contain a von Willebrand Adomain that is crucial for toxin binding. Here we report that integ-rin β1, another cell surface von Willebrand A domain protein, canmediate and potentiate anthrax toxin endocytosis. By using micro-array-based analysis to globally correlate gene expression profileswith toxin sensitivity, we associated toxin effectswith the integrin-activating proteins osteopontin and CD44. Further study showedthat PA binds to α4β1– and α5β1–integrin complexes, leading totheir conjoint endocytosis, and also interacts—weakly relative toCMG2 but comparably to TEM8—with purified α5β1 complex invitro. Monoclonal antibody directed against β1-integrin or its αintegrin partners reduced PA/integrin endocytosis and anthraxtoxin lethality, and hyaluronic acid—which interferes with CD44-mediated integrin activation—had similar effects. Remarkably,whereas deficiency of CMG2 protected macrophages from rapidkilling by large toxin doses (>50 ng/mL), by 24 h the toxin-treatedcells were dead. Such late killing of CMG2-deficient cells by highdose toxin as well as the late death observed during exposure ofCMG2-producing macrophages to low-dose toxin (<1 ng/mL), wasdependent on integrin function. Effects of inactivating both CMG2and integrin were synergistic. Collectively, our findings arguestrongly that β1-integrin can both potentiate CMG2-mediated en-docytosis and serve independently as a low-affinity PA receptor.

CD44 | microarray | osteopontin | CMG2 | lethal factor

Lethal factor (LF), the principal virulence factor of Bacillusanthracis, inhibits the host MAPK signaling pathway (1). To

enter cells, LF interacts with another B. anthracis protein, pro-tective antigen (PA), which in turn binds to cell surface receptorsencoded by two genes: TEM8 (2) and CMG2 (3). Entry of thePA-LF complex is modulated by the GTPase-activating protein,ARAP3 (4) and LDL-related protein 6 (LRP6) (5), as well as byother genes that affect autophagy (6) or clathrin-mediated en-docytosis (7, 8). Whereas the mechanisms underlying the effectsof ARAP3 on anthrax toxin entry are not known, LRP6 has beenshown to interact with TEM8 and CMG2 at the cell surface (5)and to accelerate their endocytosis (9).Both TEM8 and CMG2 are type I membrane proteins con-

taining a vonWillebrand factor A (vWA) domain (2, 3, 10), whichwas identified originally in a serum protein important for the ad-hesion of blood platelets (11). Although the normal physiologicalroles of TEM8 and CMG2 are unknown, both receptor proteinsbind to at least some of their ligands, including B. anthracis PA, byinteracting through its cation-dependent metal ion–dependentadhesion site (MIDAS) within the vWA domain (10, 12). Recentevidence indicates that genetic inactivation of CMG2 in mice hasprofound effects on anthrax toxin lethality, whereas TEM8 in-activation has little effect (13). The CMG2 and TEM8 receptorsare differentially expressed by different types of cells (2, 3, 10).

The vWA domain is also known as an “integrin-like domain”because of its occurrence in multiple, but not all, integrins, andthe binding of TEM8 and CMG2 to anthrax toxin has been com-pared with attachment of integrin to ligands (10, 12). Integrinsare a family of cell surface adhesion proteins that mediate in-teractions among cells or between cells and the extracellularmatrix (14). Functional integrin complexes are formed by thejoining of α and β subunits in the endoplasmic reticulum to formheterodimers, which are then activated by a conformationalchange that exposes sites involved in ligand binding (14). Dif-ferent cells express different integrin complexes, which carry outdisparate biological functions (14). The ligand binding sites ofintegrin complexes can be blocked highly specifically by mono-clonal antibodies (15–19).Previous investigations from our laboratory have used phenotype-

based assays to identify host cell proteins that affect the internal-ization of PA (4, 5). By using a microarray-based bioinformaticsapproach, we identified integrin-related genes whose expression cor-relates with sensitivity to LF-PA in multiple cell lines. This approachhas led to discovery of the role of the αβ integrin complex in anthraxtoxin endocytosis.

ResultsIdentification of Genes that Are Differentially Expressed in LF-PA–Sensitive Versus LF-PA–Resistant Cell Lines. Diverse cell lines trea-ted with LF-PA show a wide range of sensitivities to the toxin (9,20–25), and cell death occurs only in cells that have impairedMAPK kinase function (21). We hypothesized that such pheno-typic differences might enable microarray-based identification ofother genes whose differentially elevated or reduced expressioncorrelates with differential toxin lethality. Using pattern-searchalgorithms of GABRIEL (Genetic Analysis By Rules Incor-poratingExpert Logic), a rule-based systemof computer programsdesigned for genetic analyses (26), to detect such correlations ina previously published dataset of gene expression profiles for NCI60 tumor cell lines (27), we identified nine genes whose reducedexpression in LF-PA–resistant versus LF-PA–sensitive cell linesexceeded the variation in global gene expression in those cell linesby at least twofold (Fig. S1). The false-positive rate for thisthreshold choice was 0.09. Using the same search parameters, nogenes were found by GABRIEL to be up-regulated in LF-PA–

resistant and down-regulated inLF-PA–sensitiveNCI 60 cell lines.Quantitative RT-PCR analysis of mRNA from two additional

cell lines found previously to be highly sensitive [RAW264.7mousemacrophages (28)] or highly resistant [M2182 human prostate

Author contributions: M.M., S.-Y.J., and S.N.C. designed research; M.M. and S.-Y.J. per-formed research; M.M., S.-Y.J., and S.N.C. analyzed data; and M.M. and S.N.C. wrote thepaper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: sncohen@stanford.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1010145107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1010145107 PNAS | August 31, 2010 | vol. 107 | no. 35 | 15583–15588

MICRO

BIOLO

GY

cancer cells (4, 5)] to LF-PA showed that expression of three ofthe nine GABRIEL-detected genes, osteopontin (OPN), DAB2,and cystatin B, in RAW264.7 cells exceeded expression in M2182cells by 350,000-, 240-, and 80-fold, respectively, consistent withpossible association of gene expression with anthrax toxicity; sta-tistically significant differential expression of the other six geneswas not observed. OPN is a secreted phosphoprotein known toactivate three distinct integrin complexes—α4β1,α5β1, and αvβ3—through its interaction with the cell surface protein CD44 (29–34).Importantly, the β components of these integrin complexes containthe vWAdomain, which has been shown to be a site for PA bindingin the two previously identified anthrax toxin receptors. Together,these considerations raised the possibility that integrins β1 and/orβ3, like TEM8 and CMG2, may function as receptors for PA. Thefollowing results were obtained in experiments designed to testthis notion.

α4β1 and α5β1-integrin Complexes Colocalize with PA at the Macro-phage Cell Surface and During Endocytosis. Fluorescence micros-copy of RAW264.7 macrophages exposed to PA and anti–β1-integrin antibodies that have been labeled differentially withfluorescent dyes showed colocalization of PA with integrin at thesurface of cells maintained at 4 °C but no detectable integrin orPA in the cytoplasm (Fig. 1A). Shift of the cells to 37 °C, whichleads to PA uptake (4, 5), resulted in the conjoint internalizationof β1-integrin and PA and their continued colocalization in thecytoplasm. However, in the absence of PA, β1-integrin remainedat the surface after shift of macrophages to 37 °C, indicatingthat its uptake was dependent on PA. This finding, together withearlier evidence that integrins routinely undergo endocytosisupon binding to specific ligands (35), argues that PA and β1-integrin are constituents of a PA internalization complex. Sup-porting this conclusion, fluorescence microscopy using differen-tially labeled PA and antibodies to integrin α4 or α5, which arebinding partners of β1-integrin, showed that these α integrins alsocolocalized with PA at the cell surface, and upon uptake of PA at37 °C, also in the cytoplasm (Fig. 1B andC).Macrophages treatedwith antibodies to integrin β3 showed no corresponding fluores-cence, consistent with earlier investigations showing that thisintegrin is not expressed in RAW264.7 macrophages (36, 37).Unexpectedly, given the concurrent presence on the surface of

RAW264.7 cells of CMG2 (38, 39), which has been describedas the major PA receptor (13), prior treatment of cells with amonoclonal anti-β1 antibody known to highly specifically blockthe interaction of this integrin with its ligands (18) decreasedthe binding of PA to RAW264.7 cells by at least 60% as indicatedby Western blotting, and also reduced the uptake of PA intothe cytoplasm. In contrast, antibody against β3 integrin, which isnot produced by these macrophages, did not detectably affecteither the attachment (Fig. 2 A and B) or internalization (Fig. 2 A

and C) of PA. These results suggest that, in RAW264.7 cells,integrin β1 affects the actions of CMG2 as a PA receptor.

Biochemical Evidence of Interaction of α5β1 Complex with PA. Bio-chemical evidence of the ability of the purified β1-integrin–containing complex α5β1 to bind to PA in vitro, and the dissoci-ation constant for the interaction, were obtained using a surfaceplasmon resonance (SPR)-based optical biosensor (ProteOnXPR36; Bio-Rad). The equilibrium dissociation constant (Kd) wedetermined for the PA/integrin complex, 1 μM, approximates thatobserved for the PA/TEM8 interaction using SPR methods (10),but both values are much lower than the Kd calculated from dataobtained using a cell-based assay of the effects of PA inhibitorson toxin lethality (13). We found no interaction between α5β1-integrin and PA in the absence of divalent cation (Fig. S2), stronglyarguing that the interaction is mediated by the vWA/MIDAS do-main of β1-integrin. The purified complex of β1-integrin with α4-integrin was not tested.

Functional Role of α-Integrin Components of β1-integrin Complexesin RAW264.7 Macrophage Sensitivity to LF-PA. The ability β1-integrin to bind to ligands depends on its prior interaction witha suitable α integrin partner (14), which can be either α4 or α5integrin—neither of which contain a vWA domain (14). Mono-clonal antibodies directed against specific epitopes in α integrinscan prevent this interaction (15, 16). We found that treatment ofRAW264.7 macrophages with such anti-α4 (16) or anti-α5 (15)integrin antibodies increased survival of macrophages exposed toLF-PA (Fig. 3A)—providing further confirmation of the role ofintegrins in PA-dependent anthrax toxin lethality and establish-ing the role of αβ integrin complex formation in this process.Similarly, the purified α5β1-integrin complex also increased theability of RAW264.7 macrophages to survive the lethality of LF-PA (Fig. 3B) suggesting that free αβ integrin complexes may beable to competitively inhibit the action of such complexes at thecell surface. Monoclonal antibodies that inhibit the formation ofαβ complexes containing integrin αv (17) or β3 (19) had no de-tectable effect on the ability of these macrophages to survive ex-posure to LF-PA (Fig. 3A), consistent with the absence ofdetectable β3 integrin on the surface of RAW264.7 macrophages(Fig. 2) (36, 37) and the absence of αv in integrin complexescontaining β1. The anti-α5 integrin monoclonal antibody we usedhas a relatively low affinity for the ligand-binding site on its tar-geted integrin (http://www.biolegend.com) and showed a corre-spondingly small effect on anthrax toxicity. Control experimentsindicated that none of the antibodies we tested detectably affectedmacrophage survival in the absence of the toxin (Fig. S3)

Effects of CD44 and Hyaluronic Acid on LF-PA Lethality. The cellsurface protein CD44 participates with OPN in the activation ofintegrins and increases their binding to ligands (30–32). Fluo-

Fig. 1. Localization of β1, α4, and α5 integrins during PA binding and endocytosis in RAW264.7 cells. Fluorescence microscopy images showing the cellularlocalization of fluorescently labeled PA (green) and fluorescently labeled antiintegrin antibodies (red). (A) β1-Integrin, (B) α4-integrin, and (C) α5-integrinlocalization during PA binding (4 °C) and PA internalization (37 °C). Integrin internalization was also measured in the absence of PA.

15584 | www.pnas.org/cgi/doi/10.1073/pnas.1010145107 Martchenko et al.

rescence microscopy indicated that CD44 colocalizes with PAupon binding to the cell surface (Fig. S4), as do integrins, butunlike integrins, CD44 was not detectably endocytosed along withPA. Additionally, as shown in Fig. 3A, a monoclonal antibody thatblocks the interaction between CD44 and OPN (41) protectedmacrophages against late killing by LF-PA without affecting cellsurvival in the absence of the toxin (Fig. S3). These results providestill further evidence to support the role of integrins in PA in-ternalization and toxicity, and additionally argue that integrinactivation is required for its PA-internalizing function.The integrin-activating function of the CD44/OPN complex is

antagonized by hyaluronic acid (HA), an anionic glycosamino-glycan (40) that interacts with CD44 and largely negates theintegrin-activating effects of OPN and CD44. As we observedthat integrin activation (40) is necessary for anthrax toxicity, wetested the effects of high molecular weight HA from human

umbilical cord [estimated molecular weight of 3,420,000 Da (42)]on internalization of PA and on the lethality of the PA-LFcomplex. Addition of 100 μg/mL HA to cultures of RAW264.7 macrophages reduced PA binding and PA entry into thecytoplasm by 50% to 70% as shown by fluorescence microscopyand Western blotting (Fig. 4 A–C), and addition of HA alsoreduced β1-integrin–PA cytoplasmic colocalization during toxininternalization (Fig. 4D and Fig. S5)—consistent with the pre-viously reported ability of HA to interfere with CD44-mediatedintegrin activation. However, addition of HA had no detectableeffect on survival of macrophages treated with more than 50 ng/mL LF (Fig. 4E), suggesting that notwithstanding the modestdecrease in binding and internalization of toxin resulting fromantiintegrin measures, this toxin dose achieved sufficient inter-nalization to achieve cell death. However, as was observed forantiintegrin antibodies, HA treatment at 100 μg/mL dramaticallyenhanced the ability of macrophages to survive exposure to toxin,increasing the IC50 by 32-fold (from 0.12 ng/mL to 3.85 ng/mL)and higher concentrations of HA did not further affect LF-PAlethality (Fig. S6). Low molecular weight HA degradation prod-ucts have been reported to not bind to CD44 (43), and we foundthat HA products of Streptococcus pyogenes fermentation (900,000Da, 132,000 Da, 16,100 Da, and 6,400 Da; R&D Systems), did notdetectably change the LF-PA sensitivity of RAW264.7 cells.Combined use of anti–β1-integrin monoclonal antibody with

100 μg/mL of HA did not result in any further reduction in cel-lular sensitivity to the toxin (Fig. 4E), supporting the notion thatthese agents target the same pathway; similarly, the effects of HAand the knockdown of expression of ARAP3, which previouslyhas been implicated in PA entry by an unknown mechanism(4), were nonadditive (Fig. 5A). In contrast, however, siRNA-induced deficiency of LRP6, which complexes with CMG2 at thecell surface and colocalizes with that receptor, reduced macro-phage sensitivity to LF-PA beyond what was observed at the

Fig. 3. Effect of antiintegrin inhibitory monoclonal antibodies on an-thrax LF-PA–mediated lethality in RAW264.7 macrophages. (A) Effect of in-hibition of antibody to integrin subunits α4, α5, αv, β1, and β3 or to theintegrin-activating protein CD44. RAW264.7 cells were incubated for 24 hwith 3 ng/mL PA and 500 ng/mL LF and in the presence of increasing amountsof purified monoclonal antibodies (0–10 μg/mL). (B) Soluble α5β1-integrinprotein inhibits the LF-PA intoxication of RAW264.7 cells. RAW264.7 cells thatexpress endogenous anthrax toxin receptors were incubatedwith 3 ng/mL PAand 500 ng/mL LF and in the presence of increasing amounts of purified α5β1-integrin protein (0–31.75 μg/mL) for 24 h.

Fig. 4. Effects of CD44 and HA on the sensitivity of RAW264.7 macro-phages to LF-PA. (A) Fluorescence microscopy shows effects of HA on PAinternalization. Western blot analysis of the influence of HA on the in-ternalization of monomeric PA in the cytoplasm (B), and on heptameric PAbinding to the cell surface (C). Numbers represent the average of relativeintensities of PA ± SD obtained from three independent experiments. (D)Effect of HA on colocalization between β1-integrin and PA. Fluorescencemicroscopy images showing the colocalization of fluorescently labeled anti–β1-integrin antibodies (red) and PA oligomers (green). (E) Effects of exoge-nously addedHAon susceptibility of RAW264.7 cells to LF-PA toxin. RAW264.7cells were plated in the presence or in the absence of HA (100 μg/mL), anti–β1-integrin inhibitory antibody, or both in 96-well plates and were treated withserially diluted LF in the presence of constant of PA (200 ng/mL) for 24 h.Numbers represent IC50 values.

Fig. 2. Integrin/PA interaction. (A) Fluorescence microscopy analysis of theeffects of inhibitory monoclonal anti–β1-integrin and anti–β3 antibodies onPA internalization. RAW264.7 cells were incubated with Alexa Fluor 488–labeled PA (1 μg/mL) in the presence or in the absence of anti-β1 or anti–β3integrin antibody for 20 min at 37 °C followed by fluorescence microscopy.Western blot analysis of the influence of anti-β1 and -β3 integrin inhibitoryantibody on monomeric PA binding to the cell surface (B), and on the in-ternalization of heptameric PA in the cytoplasm (C). Numbers represent theaverage of relative intensities of PA ± SD obtained from three independentexperiments.

Martchenko et al. PNAS | August 31, 2010 | vol. 107 | no. 35 | 15585

MICRO

BIOLO

GY

maximally effective concentration of HA (IC50 of 17.4 in LRP6-deficient cells vs. IC50 of 3.08 in parental cells; Fig. 5B).

Integrin Effects on Early and Late LF-Mediated Killing of Macro-phages. RAW264.7 macrophages have been reported to undergoLF-mediated killing by two distinct mechanisms: a rapid lyticdeath occurs in the presence of high LF concentrations, whereasa slow caspase-dependent apoptotic death is observed at lowLF concentrations (44, 45). A recent publication indicates thatCMG2 is necessary for the lethality of high doses of anthraxtoxin in mice and MEFs derived from them (13). Using shRNAthat reduced the steady-state level of CMG2 expression inRAW264.7 macrophages by 70%, we confirmed that CMG2knockdown reduces the early death occurring after 4 h of exposureto more than 50 ng/mL LF (Fig. 5C). However, the same knock-down of CMG2 did not affect late death, as assayed by survivalafter 24-h exposure to the same LF concentration (Fig. 5D), im-plying that either the CMG2 expression remaining in these cellsenabled entry of sufficient toxin to achieve late death, or alter-natively, that late death can occur by a CMG2-independentmechanism. Only late death was observed in macrophages treatedwith a low concentration of LF (i.e., <1 ng/mL), as reported pre-viously by Popov et al. (44), and CMG2 deficiency had no de-tectable effect on this lethality (Fig. 5D), arguing that late killing bylow-dose toxin in macrophages is independent of CMG2. In-terestingly, whereas the HA-mediated interference with integrinhad no detectable effect on early macrophage death (Fig. 5C), itfurther reduced the lethality to CMG2-deficent cells. The findingis consistent with our earlier evidence (Fig. 2) that antibody to

integrin β1 reduces the ability of CMG2 to act as a PA receptor inRAW264.7 macrophages.

DiscussionThe results reported here indicate a dual role for integrins inthe entry of anthrax toxin into mouse macrophages. The abilityof integrin complexes to respond to PA binding by conjointly en-tering the cytoplasm with PA, plus the finding that purified inte-grin α5β1 complexes can interact with purified PA with appro-ximately the same affinity as TEM8 (10), suggests that integrinscan function as a PA receptor. Like both of the previously iden-tified anthrax receptors, TEM8 and CMG2, β1-integrin includesa PA-binding vWA domain containing the MIDAS motif, whichis absolutely conserved in cation-dependent ligand binding ofintegrins (14). Structural analysis indicate that the vWAdomains ofCMG2 and integrin fold similarly (46) and PA–CMG2 complexesshow mimicry of integrin–ligand interactions (12, 47). However,the finding that integrin inactivation by antibodies or HA reducesthe ability of concurrently present CMG2 to internalize PA andmediate lethality in RAW264.7 macrophages argues that integrinscan also act as modulators of CMG2 receptor function.Two distinct mechanism of killing of macrophages exposed to

LF have been reported (28, 44), and our investigations provideadditional evidence of this. Macrophages treated with high dosesof LF undergo rapid cell lysis and death (28, 44), and—notsurprisingly, given similar results obtained for mouse embryonicfibroblasts (13)—we show here that macrophage death occurringsoon after exposure to the toxin is dependent on CMG2 function(Fig. 5C). HA-mediated interference with integrin had no de-tectable effect on CMG2-dependent early macrophage death

Fig. 5. Establishment of pathways involved in LF-PA entry. Effects of exogenously added HA on susceptibility of WT RAW264.7 cells: ARAP3 knockdown (A),LRP6 knockdown (B), or CMG2 knockdown (D) cells to 24 h exposure to LF-PA toxin. (C) Effects of exogenously added HA on susceptibility of WT RAW264.7cells or CMG2 knockdown cells to 4 h exposure to LF-PA toxin. Numbers represent IC50 values. The analysis was done as in Fig. 4E.

15586 | www.pnas.org/cgi/doi/10.1073/pnas.1010145107 Martchenko et al.

(Fig. 5C). However, RAW264.7 macrophages defective in CMG2expression nevertheless underwent eventual killing by toxin (Fig.5D), implying that a second mechanism exists for toxin entry inthese cells. This mechanism, which necessarily cannot involveTEM8 [which is not present on the surface of macrophages (38,39)] was inhibited by HA (Fig. 5D). Late apoptotic killing alsooccurs after exposure of macrophages to sublytic doses of LF(refs. 28, 44 and our data; Fig. 5D), and such killing, although notaffected by an shRNA-mediated decrease in CMG2 expression,was decreased byHAaswell as by antibodies specific to β1-integrinand its α integrin partners. It is worth noting that integrins alsomediate the lethality of the α toxin of Staphylococcus aureus (48,49), and that such killing by lytic or apoptotic mechanisms alsoappears to be concentration-dependent (50).The effects of maximal interference with integrin by HA are

additive to those of knockdown of LRP6, a previously identifiedmember of an anthrax toxin internalization complex (5, 9), whereasHA is nonadditive to knockdown of ARAP3, another modulatorof PA entry (Fig. 5 A and B). These findings suggest that ARAP3and LRP6 affect different pathways of toxin internalization andthat the ARAP3 pathway is at least partially congruent with thepathway of integrin-mediated internalization. During these ex-periments, we observed, surprisingly, that interference with dif-ferent entry-modulating proteins has differential effects on thedeath of RAW264.7 macrophages induced by exposure to PA plusLF versus PA plus FP59, a hybrid toxin containing the PA bindingsite of LF plus a toxin domain derived from Pseudomonas aerugi-nosa exotoxin A.The synergistic—rather than simply additive—effects of HA

and shRNA directed against CMG2 on late death by sublytic LFconcentrations suggest that integrin may enhance action of re-sidual CMG2 remaining after shRNA knockdown, further sup-porting the notion that integrins can facilitate the receptor func-tions of CMG2. Similarly, TEM8 may also potentiate CMG2knockout in addition to serving as a low-affinity receptor (13).TEM8 has been reported to cooperate with integrin in regulatingthe VEGFR2 receptor (51), and analogous cooperation betweenTEM8 and integrins may occur in regulation of CMG2-mediatedtoxin entry.In animals, C57BL/6 mice containing null mutations in TEM8

and CMG2 survive toxin exposure (13), suggesting that integrinfunction, which was not manipulated in the tested mice, wasinsufficient for LF lethality in the experimental system used. Thismay be attributable to the known differential expression of an-thrax receptors among different host cells and tissues (2, 3, 10),and potentially may result in part from a lack of relevant integrinson the surface of the yet-unidentified cells that mediate death ofmice exposed to LF. Whereas CMG2 is necessary for normalkilling of C57BL/6 mice, CMG2-null mutants were only partiallyprotected from the effects of LF injections (13), again suggestingthe existence of an additional functional receptor, which wasshown in these animals to be TEM8 (13). However, the relevanceof mouse models that use high-dose toxin injections or massiveamounts of spores (13) to the natural process of anthrax infectioncan reasonably be questioned: rhesus monkey models that moreclosely resemble the conditions of human infection by B. anthracishave been observed to yield relatively low serum levels of toxinthat are detected as a relatively late event (52, 53).

Materials and MethodsBioinformatics Analysis of Gene Expression. GABRIEL software (26) separatedcancer cell lines (SK-MEL-5, SK-MEL28, M-14, MALME-3M, MDA-MB-435,MDA-N, SK-MEL-2, HL-60, MCF7, SW-620, EKVX, and A549) into two groups,LF-PA–sensitive and LF-PA–resistant, based on previous reports (9, 20–25).GABRIEL (27) was used for correlation of steady-state levels of expressionwith toxin lethality.

Chemicals and Reagents. PA and LF were purchased from List BiologicalLaboratories. The fluorescently labeled antibodies purchased from Bio-Legend were anti–CD44-APC, anti–β1-integrin-APC, anti–α4-integrin/AlexaFluor 647, and anti–α5 integrin/Alexa Fluor 647. The preservative-freemonoclonal antibodies purchased from BioLegend were anti-CD44 (IM7),anti–β1-integrin (HMβ1-1), anti–α4-integrin (R1-2), anti–α5 integrin (MFR5),anti–β3 integrin (HMβ3-1), and anti–αv integrin (RMV-7). High molecularweight HA from human umbilical cord was purchased from Sigma-Aldrichand HA from S. pyogenes of high, medium, low, and ultra-low molecularweight were purchased from R&D Systems.

Cell Culture. RAW264.7 mouse macrophage cells were maintained in DMEM(Invitrogen) supplemented with 10% FBS (HyClone) and 100 μg/mL penicillinand 100 μg/mL streptomycin. In this study we isolated single clones froma pool of ARAP3, LRP6 knockdown, and their respective WT RAW264.7 celllines, which were described previously (4, 5). For CMG2 knockdown experi-ment, lentiviral-based shRNAmir vector pGIPZ with sequence targeted forCMG2 was purchased from Open Biosystems. Viral infection of CMG2 shRNAwas performed using lentiviral-based methods (4, 5). Single RAW264.7 col-onies were selected after puromycin treatment (4 μg/mL).

Toxin Treatment and Cell Viability Assays. Cells were treated with toxin for 4 or24 h, and determination of cell viability by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was performed as described (4, 5).Each data point shown in figures for MTT assays represents the average ofresults from four wells in each of at least three separate experiments. Cellviability is shown as the percentage of survivors obtained relative to treat-ment by PA alone (100%).

Antibody/Integrin Protection Assay. RAW264.7 cells (3 × 103 per well) wereseeded in 96-well plates (100 μL/well) 24 h before the assay. The cells werethen incubated at 37 °C for 24 h with 3 ng/mL of PA, 500 ng/mL of LF, andserial dilutions of antibodies or purified α5β1-integrin (R&D Systems), fol-lowed by 24-h incubation in fresh, toxin-free media. The preservative-freemonoclonal antibodies were suspended in the media at various concen-trations. Toxin sensitivity was determined by MTT assay.

Immunofluorescence Microscopy. PA protein was labeled with Alexa Fluor488 using the A10235 protein labeling kit (Molecular Probes). We determinedby MTT assay that such labeling did not affect the ability of PA-LF to killmacrophages. For experiments determining the effects of β1 and β3 in-tegrin inhibition or HA on toxin cell entry, cells were preincubated with100 μg/mL HA or with 10 μg/mL of antimouse β1 or β3 integrin inhibitorymonoclonal antibodies for 1 h in serum-free Iscove modified Dulbecco me-dium (IMDM; Invitrogen). Fresh serum-free IMDM media was added thatcontained 1 μg/mL PA/Alexa Fluor 488 alone or 1 μg/mL PA/Alexa Fluor488 mixed with fresh 100 μg/mL HA or with 10 μg/mL of antimouse β1/β3integrin inhibitory monoclonal antibodies. Cells were incubated for 60 minat 4 °C for PA-binding analysis and for 20 min at 37 °C for determination ofPA internalization.

For localization of integrins and CD44, cells were incubated with orwithout 1 μg/mL PA/Alexa Fluor 488, and with 0.5 μg/mL fluorescently la-beled antibodies for 60 min at 4 °C for PA-binding analysis and for 20 min at37 °C for PA-internalization analysis in serum-free IMDM media (Invitrogen).Cells were washed, fixed, and examined by using a fluorescence micro-scope (Leica).

Binding Kinetics. Experiments were performed using the Bio-Rad ProteOnXPR36 system. The extracellular portions of α5 and β1-integrins, devoidof their transmembrane and cytoplasmic domains, were cloned up-stream of acidic and basic tails, and the heterodimer was expressed inand purified from CHO cell line by R&D Systems. The α5β1 heterodimerwas functional based on its ability to bind fibronectin with Kd ofnanomolar range (R&D Systems). Protein concentration of α5β1-integrinwas 0.2 μmol/L and the concentrations of PA ranged from 148 nmol/L to12 μmol/L in PBS buffer with 1 mmol/L MnCl2. Equilibrium dissociationconstants were calculated on the basis of the kinetic measurements ofthe association and dissociation rate constants according to the formulaKd = kd / ka.

Biochemical Assay of PA Binding and Internalization. Cells were grown toconfluence and preincubated in serum-free IMDM media (Invitrogen)with 10 μg/mL of inhibitory monoclonal antibodies in β1 or β3 integrins orwith 100 μg/mL of HA for 1 h before LF-PA exposure. Cells were exposed to

Martchenko et al. PNAS | August 31, 2010 | vol. 107 | no. 35 | 15587

MICRO

BIOLO

GY

1 μg/mL of PA at 4 °C for 1 h for binding assay or 100 ng/mL of PA at 37 °Cfor 20 min for internalization assay in the presence or in the absence of HAor inhibitory antibody in serum-free IMDM media. Cells were then washedwith cold PBS solution five times and lysed in RIPA buffer containing a pro-tease inhibitor mixture (Roche). Western blot analysis was performed usingrabbit polyclonal anti-PA antibody 7349.3 (kindly provided by G. Bokoch,Scripps Research Institute, La Jolla, CA) to detect monomeric PA, rab-bit polyclonal anti-PA antibody (Abcam) to detect heptameric PA, ormouse antiactin monoclonal antibody (Sigma-Aldrich). Chemiluminescence

of bands and their relative intensities were revealed using a VersaDoc 1000instrument (Bio-Rad).

ACKNOWLEDGMENTS. We thank Richard Lin for bioinformatics assistance,BioRad for assistance with and reagents for biochemical binding experi-ments, and Dr. G. Van Der Goot for comments on the manuscript. M.M.received a postdoctoral fellowship from the Fonds de la recherche en santédu Québec. This study was supported by Defense Threat Reduction AgencyAward HDTRA1-06-C-0039 (to S.N.C.).

1. Duesbery NS, et al. (1998) Proteolytic inactivation of MAP-kinase-kinase by anthraxlethal factor. Science 280:734–737.

2. Bradley KA, Mogridge J, Mourez M, Collier RJ, Young JA (2001) Identification of thecellular receptor for anthrax toxin. Nature 414:225–229.

3. Scobie HM, Rainey GJ, Bradley KA, Young JA (2003) Human capillary morphogenesisprotein 2 functions as an anthrax toxin receptor. Proc Natl Acad Sci USA 100:5170–5174.

4. Lu Q, Wei W, Kowalski PE, Chang AC, Cohen SN (2004) EST-based genome-wide geneinactivation identifies ARAP3 as a host protein affecting cellular susceptibility toanthrax toxin. Proc Natl Acad Sci USA 101:17246–17251.

5. Wei W, Lu Q, Chaudry GJ, Leppla SH, Cohen SN (2006) The LDL receptor-relatedprotein LRP6 mediates internalization and lethality of anthrax toxin. Cell 124:1141–1154.

6. Ha SD, et al. (2010) Cathepsin B-mediated autophagy flux facilitates the anthrax toxinreceptor 2-mediated delivery of anthrax lethal factor into the cytoplasm. J Biol Chem285:2120–2129.

7. Abrami L, Kunz B, van der Goot FG (2010) Anthrax toxin triggers the activation of src-like kinases to mediate its own uptake. Proc Natl Acad Sci USA 107:1420–1424.

8. Abrami L, Liu S, Cosson P, Leppla SH, van der Goot FG (2003) Anthrax toxin triggersendocytosis of its receptor via a lipid raft-mediated clathrin-dependent process. J CellBiol 160:321–328.

9. Abrami L, et al. (2008) Functional interactions between anthrax toxin receptors andthe WNT signalling protein LRP6. Cell Microbiol 10:2509–2519.

10. Scobie HM, et al. (2005) A soluble receptor decoy protects rats against anthrax lethaltoxin challenge. J Infect Dis 192:1047–1051.

11. Turitto VT, Weiss HJ, Zimmerman TS, Sussman II (1985) Factor VIII/von Willebrandfactor in subendothelium mediates platelet adhesion. Blood 65:823–831.

12. Bradley KA, et al. (2003) Binding of anthrax toxin to its receptor is similar to alphaintegrin-ligand interactions. J Biol Chem 278:49342–49347.

13. Liu S, et al. (2009) Capillary morphogenesis protein-2 is the major receptor mediatinglethality of anthrax toxin in vivo. Proc Natl Acad Sci USA 106:12424–12429.

14. Hynes RO (2002) Integrins: Bidirectional, allosteric signaling machines. Cell 110:673–687.

15. Kinashi T, Springer TA (1994) Adhesion molecules in hematopoietic cells. Blood Cells20:25–44.

16. Lobb RR, Hemler ME (1994) The pathophysiologic role of α 4 integrins in vivo. J ClinInvest 94:1722–1728.

17. Maxfield SR, et al. (1989) Murine T cells express a cell surface receptor for multipleextracellular matrix proteins. Identification and characterization with monoclonalantibodies. J Exp Med 169:2173–2190.

18. Noto K, Kato K, Okumura K, Yagita H (1995) Identification and functional char-acterization of mouse CD29 with a mAb. Int Immunol 7:835–842.

19. Phillips DR, Charo IF, Scarborough RM (1991) GPIIb-IIIa: the responsive integrin. Cell65:359–362.

20. Abi-Habib RJ, et al. (2006) A urokinase-activated recombinant anthrax toxin isselectively cytotoxic to many human tumor cell types. Mol Cancer Ther 5:2556–2562.

21. Abi-Habib RJ, et al. (2005) BRAF status and mitogen-activated protein/extracellularsignal-regulated kinase kinase 1/2 activity indicate sensitivity of melanoma cells toanthrax lethal toxin. Mol Cancer Ther 4:1303–1310.

22. Chen KH, Liu S, Bankston LA, Liddington RC, Leppla SH (2007) Selection of anthraxtoxin protective antigen variants that discriminate between the cellular receptorsTEM8 and CMG2 and achieve targeting of tumor cells. J Biol Chem 282:9834–9845.

23. Kassam A, Der SD, Mogridge J (2005) Differentiation of human monocytic cell linesconfers susceptibility to Bacillus anthracis lethal toxin. Cell Microbiol 7:281–292.

24. Koo HM, et al. (2002) Apoptosis and melanogenesis in human melanoma cellsinduced by anthrax lethal factor inactivation of mitogen-activated protein kinasekinase. Proc Natl Acad Sci USA 99:3052–3057.

25. Liu S, et al. (2008) Matrix metalloproteinase-activated anthrax lethal toxin de-monstrates high potency in targeting tumor vasculature. J Biol Chem 283:529–540.

26. Pan KH, Lih CJ, Cohen SN (2002) Analysis of DNA microarrays using algorithms thatemploy rule-based expert knowledge. Proc Natl Acad Sci USA 99:2118–2123.

27. Ross DT, et al. (2000) Systematic variation in gene expression patterns in humancancer cell lines. Nat Genet 24:227–235.

28. Hanna PC, Acosta D, Collier RJ (1993) On the role of macrophages in anthrax. ProcNatl Acad Sci USA 90:10198–10201.

29. Bayless KJ, Meininger GA, Scholtz JM, Davis GE (1998) Osteopontin is a ligand for theα4β1 integrin. J Cell Sci 111:1165–1174.

30. Katagiri YU, et al. (1999) CD44 variants but not CD44s cooperate with β1-containingintegrins to permit cells to bind to osteopontin independently of arginine-glycine-aspartic acid, thereby stimulating cell motility and chemotaxis. Cancer Res 59:219–226.

31. Ponta H, Sherman L, Herrlich PA (2003) CD44: From adhesion molecules to signallingregulators. Nat Rev Mol Cell Biol 4:33–45.

32. Redondo-Muñoz J, et al. (2008) α4β1 integrin and 190-kDa CD44v constitute a cellsurface docking complex for gelatinase B/MMP-9 in chronic leukemic but not innormal B cells. Blood 112:169–178.

33. Scatena M, et al. (1998) NF-kappaB mediates alphavbeta3 integrin-induced endo-thelial cell survival. J Cell Biol 141:1083–1093.

34. Yokosaki Y, Tanaka K, Higashikawa F, Yamashita K, Eboshida A (2005) Distinctstructural requirements for binding of the integrins alphavbeta6, alphavbeta3,alphavbeta5, α5β1 and α9β1 to osteopontin. Matrix Biol 24:418–427.

35. Caswell PT, Vadrevu S, Norman JC (2009) Integrins: Masters and slaves of endocytictransport. Nat Rev Mol Cell Biol 10:843–853.

36. Cuetara BL, Crotti TN, O’Donoghue AJ, McHugh KP (2006) Cloning and char-acterization of osteoclast precursors from the RAW264.7 cell line. In Vitro Cell DevBiol Anim 42:182–188.

37. Hirotani H, Tuohy NA, Woo JT, Stern PH, Clipstone NA (2004) The calcineurin/nuclearfactor of activated T cells signaling pathway regulates osteoclastogenesis in RAW264.7cells. J Biol Chem 279:13984–13992.

38. Dal Molin F, et al. (2006) Cell entry and cAMP imaging of anthrax edema toxin. EMBOJ 25:5405–5413.

39. Rainey GJ, et al. (2005) Receptor-specific requirements for anthrax toxin delivery intocells. Proc Natl Acad Sci USA 102:13278–13283.

40. Knudson CB, Knudson W (2004) Hyaluronan and CD44: Modulators of chondrocytemetabolism. Clin Orthop Relat Res 427(suppl)S152–S162.

41. Mikecz K, Brennan FR, Kim JH, Glant TT (1995) Anti-CD44 treatment abrogates tissueoedema and leukocyte infiltration in murine arthritis. Nat Med 1:558–563.

42. Saari H, Konttinen YT, Friman C, Sorsa T (1993) Differential effects of reactive oxygenspecies on native synovial fluid and purified human umbilical cord hyaluronate.Inflammation 17:403–415.

43. Gariboldi S, et al. (2008) Low molecular weight hyaluronic acid increases the self-defense of skin epithelium by induction of β-defensin 2 via TLR2 and TLR4. J Immunol181:2103–2110.

44. Popov SG, et al. (2002) Lethal toxin of Bacillus anthracis causes apoptosis of macro-phages. Biochem Biophys Res Commun 293:349–355.

45. Wu W, et al. (2009) Resistance of human alveolar macrophages to Bacillus anthracislethal toxin. J Immunol 183:5799–5806.

46. Lacy DB, Wigelsworth DJ, Scobie HM, Young JA, Collier RJ (2004) Crystal structure ofthe von Willebrand factor A domain of human capillary morphogenesis protein 2: ananthrax toxin receptor. Proc Natl Acad Sci USA 101:6367–6372.

47. Santelli E, Bankston LA, Leppla SH, Liddington RC (2004) Crystal structure ofa complex between anthrax toxin and its host cell receptor. Nature 430:905–908.

48. Liang X, Ji Y (2007) Involvement of α5β1-integrin and TNF-α in Staphylococcus aureusα-toxin-induced death of epithelial cells. Cell Microbiol 9:1809–1821.

49. Liang X, Ji Y (2006) α-toxin interferes with integrin-mediated adhesion and inter-nalization of Staphylococcus aureus by epithelial cells. Cell Microbiol 8:1656–1668.

50. Jonas D, et al. (1994) Novel path to apoptosis: Small transmembrane pores created bystaphylococcal α-toxin in T lymphocytes evoke internucleosomal DNA degradation.Infect Immun 62:1304–1312.

51. Jinnin M, et al. (2008) Suppressed NFAT-dependent VEGFR1 expression and con-stitutive VEGFR2 signaling in infantile hemangioma. Nat Med 14:1236–1246.

52. Boyer AE, et al. (2009) Kinetics of lethal factor and poly-D-glutamic acid antigenemiaduring inhalation anthrax in rhesus macaques. Infect Immun 77:3432–3441.

53. Boyer AE, et al. (2007) Detection and quantification of anthrax lethal factor in serumby mass spectrometry. Anal Chem 79:8463–8470.

15588 | www.pnas.org/cgi/doi/10.1073/pnas.1010145107 Martchenko et al.