Eur. J. Biochem. 267, 5191±5197 (2000) q FEBS 2000

Identification of residues critical for toxicity in Clostridium perfringensphospholipase C, the key toxin in gas gangrene

Alberto Alape-Giro n1,2,3,*, Marietta Flores-DõÂaz1,2,*, Isabelle Guillouard4, Claire E. Naylor5, Richard W. Titball6,Alexandra Rucavado2, Bruno Lomonte2, Ajit K. Basak5, Jose M. Gutie rrez2, Stewart T. Cole4 andMonica Thelestam1

1Microbiology and Tumorbiology Center, Karolinska Institutet, Stockholm, Sweden; 2Instituto Clodomiro Picado,

Facultad de MicrobiologõÂa; 3Departamento de BioquõÂmica, Facultad de Medicina, Universidad de Costa Rica, San JoseÂ, Costa Rico;4Unite de GeÂnetique MoleÂculaire BacteÂrienne, Institut Pasteur, Paris, France; 5Department of Crystallography, Birkbeck College, Malet Street,

London; 6Defense Evaluation and Research Agency, CBD Porton Down, Salisbury, Wiltshire, UK

Clostridium perfringens phospholipase C (PLC), also called a-toxin, is the major virulence factor in the

pathogenesis of gas gangrene. The toxic activities of genetically engineered a-toxin variants harboring single

amino-acid substitutions in three loops of its C-terminal domain were studied. The substitutions were made in

aspartic acid residues which bind calcium, and tyrosine residues of the putative membrane-interacting region. The

variants D269N and D336N had less than 20% of the hemolytic activity and displayed a cytotoxic potency

103-fold lower than that of the wild-type toxin. The variants in which Tyr275, Tyr307, and Tyr331 were

substituted by Asn, Phe, or Leu had 11±73% of the hemolytic activity and exhibited a cytotoxic potency 102- to

105-fold lower than that of the wild-type toxin. The results demonstrated that the sphingomyelinase activity and

the C-terminal domain are required for myotoxicity in vivo and that the variants D269N, D336N, Y275N, Y307F,

and Y331L had less than 12% of the myotoxic activity displayed by the wild-type toxin. This work therefore

identifies residues critical for the toxic activities of C. perfringens PLC and provides new insights toward

understanding the mechanism of action of this toxin at a molecular level.

Keywords: bacterial toxins; muscular diseases; molecular models; skeletal muscle; cell survival.

Bacterial phospholipases C (PLCs) are a heterogeneous groupof virulence factors with diverse roles in the pathogenesis ofseveral diseases [1]. PLCs have been classified into threegroups according to whether the preferred substrate is PtdCho,sphingomyelin (SM) or PtdIns [2]. The PtdCho-preferringbacterial PLCs include Zn21-metalloenzymes from Bacilluscereus, Listeria monocytogenes and several Clostridium spp.[2]. They display different abilities to hydrolyse SM andshow 30±60% amino-acid sequence identity [2]. However,the PLCs from clostridia possess 115 additional residues atthe C-terminus [3]. The Clostridium perfringens PLC, whichdisplays platelet aggregating, hemolytic, cytotoxic, myotoxic,and lethal activities, is the most toxic among clostridial PLCs[1±3]. This enzyme, also called a-toxin, is the major virulencefactor in the pathogenesis of gas gangrene [4,5], and has alsobeen associated with other diseases of animals and man [6].

The three-dimensional structure of a-toxin shows two domainsjoined by a short hinge region. The N-terminal domain containsthe active site and consists of 10 tightly packed a helices, similarto the nontoxic Bacillus cereus PLC [7]. The C-terminal domainis a b-sandwich composed of two four-stranded sheets [7],

analogous to C2 domains of intracellular eukaryotic proteinsinvolved in vesicular transport and signal transduction [8]. TheC-terminal domain is needed for the Ca21-dependent inter-action with aggregated PtdCho [9], and for the disruption ofartificial membranes [10], but its function is not completelyunderstood.

Crystallographic studies have revealed that a-toxin mightexist in two conformations: an open form with the active siteaccessible, and a closed form with the active site covered by theloop encompassing residues 135±150 [11]. It is suggested thatthe binding of the C-terminal domain to the target membranecauses the uncovering of the active site, thus allowing turnoverof the phospholipid substrate [11]. The Ca21-binding region ofa-toxin is located in three loops of the C-terminal domainwhich encompass residues 265±275, 292±303 and 330±339[11]. Residues surrounding the Ca21-binding ligands at the tipsof those loops are thought to constitute a membrane-interactingsurface [7]. This surface has been hypothesized to be importantfor toxicity [7], but its role in the toxic activities of a-toxin hasnot been experimentally tested.

The N-terminal domain of a-toxin, which, isolated, retainsthe lecithinase activity but lacks the sphingomyelinase activity,is not hemolytic [12] and has a cytotoxic potency 106-foldlower than that of the holotoxin [13]. Furthermore, theC-terminal domain, which, isolated, is devoid of both enzym-atic activities of the holotoxin, is neither hemolytic norcytotoxic [12,13]. Thus the interaction between the domainsis required to confer full hemolytic and cytotoxic activities toa-toxin.

Although the PtdCho-preferring PLCs from different Clos-tridium spp. are highly similar in amino-acid sequence, they

Correspondence to M. Thelestam, Microbiology and Tumorbiology Center,

Karolinska Institutet, Box 280, S-171 77 Stockholm, Sweden.

Fax: 1 46 8 33 1547, Tel.: 1 46 8 728 7162,

E-mail: monica.thelestam@mtc.ki.se

Abbreviations: PLC, phospholipase C; SM, sphingomyelin; CK, creatine

kinase.

Enzymes: phospholipase C (EC 3.1.4.3); creatine kinase (EC 2.7.3.2).

*Note: these authors contributed equally to the work.

(Received 13 April 2000, revised 16 June 2000, accepted 19 June 2000)

display strikingly different toxicities [3]. The Clostridiumbifermentans PLC, which also shows a Ca21-dependent capa-city to disrupt liposomes, has a hemolytic activity 102-foldlower and has a cytotoxic potency 105-fold lower than thatof a-toxin [13,14]. However, a hybrid protein containing theN-terminal domain of the C. bifermentans PLC and theC-terminal domain of a-toxin displays hemolytic and cytotoxicactivities significantly higher than those of the C. bifermentansPLC [13,14]. Chemical modification of the exposed tyrosineresidues in a-toxin did not affect its enzymatic activity towardp-nitrophenylphosphorylcholine, but resulted in a significantreduction of its platelet aggregating, hemolytic and lethalactivities, suggesting a role for tyrosine residues in toxicity[15]. Of the six tyrosine residues in the C-terminal domain ofa-toxin only three are conserved in the C. bifermentans PLC[9,14]. Thus, the differences in the tyrosine residues of theC-terminal regions between these two enzymes are suggested tobe at least partially responsible for the reduced toxicity of theC. bifermentans PLC [9,14].

The substitution of Asp269, Asp336, Tyr275, Tyr307 andTyr331 in a-toxin causes, at most, a 20% reduction in itsenzymatic activity toward aggregated PtdCho [9], but theimportance of these residues in toxicity has not beendetermined. The present work was conducted to clarify therole of those residues in the hemolytic, cytotoxic and myotoxicactivities.

M A T E R I A L S A N D M E T H O D S

Production and purification of toxins

The wild-type C. bifermentans PLC was expressed in Escheri-chia coli and purified [14]. The a-toxin from the strain NCTC8237 and a truncated version containing only the 249N-terminal residues were produced and purified [12,16]. Thewild-type a-toxin from strain 8-6 and the variant D56N wereproduced and purified [17,18]. The a-toxin variants D275N,D336N, Y275F, Y275N, Y307F, Y331F, Y331L, and Y367Fwere generated by site-directed mutagenesis and all mutantswere fully sequenced to verify that only the desired mutationswere present [9]. These variants were produced in E. coli strainK38/pGP1-2 using the T7 polymerase expression system andpurified from periplasmic extracts [9]. Homogeneity wasverified by SDS/PAGE and protein concentrations weredetermined using the Bio-Rad DC protein assay and BSA asa standard.

Antibodies and ELISAs

The mAbs against a-toxin 3A4D10, 3A4F2, DY2F5, 9F3A6,A5A11 were produced in pristane-primed mice and purified[19,20]. Immulon II microtiter plates were coated overnightwith 0.5 mg of each variant in 0.05 m carbonate buffer, pH 9.6.The remaining binding sites were blocked with 2% BSA inNaCl/Pi (blocking solution) for 30 min. After aspiration theantibodies serially diluted in blocking solution were addedand incubated for 1 h at 37 8C. After five washes withNaCl/Pi /0.05% Tween 20, 100 mL of anti-mouse IgG±alkalinephosphatase conjugate (Dako, Copenhagen, Denmark) wereadded per well. After a 1-h incubation at 37 8C, plates werewashed five times as above, then 100 mL of p-nitrophenyl-phosphate substrate solution (1 mg´mL21) diluted in 1 mdiethanolamine buffer, pH 9.8, was added per well, andabsorbances at 405 nm recorded after 30 min.

Molecular modeling

Amino-acid substitutions were introduced into the structureof wild-type C. perfringens a-toxin [7] using the graphicspackage, o [21]. The models were subsequently subjected torestrained energy minimization utilizing the xplor package[22], but no change in protein conformation was observed, oneither a global or a local level. Therefore, the structuresproduced by `o', prior to energy minimization were used.Figures of the models were drawn by bobscript [23] andrendered with raster 3d [24]. Surface potentials werecalculated with grasp [25].

Hemolytic, cytotoxic and myotoxic activities

The hemolytic activity was determined in agarose gels con-taining 5% sheep erythrocytes, 0.2 m sodium borate, 50 mmNaCl, 30 mm CaCl2, at pH 7.6, as described [16]. Briefly,different amounts of wild-type a-toxin in 15 mL of NaCl/Pi

were placed in 4-mm wells and after 24 h incubation in a humidchamber at 37 8C the diameter of hemolysis was measured. Alinear dose±response curve was obtained using 0.1±0.7 mg ofwild-type a-toxin and therefore 0.35 mg of each variant wasused. All experiments were performed three times, withduplicate samples each time.

The Chinese hamster cell line Don Q [26], was cultivated inpolystyrene 96-well plates (Nunc, Roskilde, Denmark) inEagle's minimal essential medium (Life Technologies, Paisley,UK) supplemented with 10% fetal bovine serum, 5 mml-glutamine, penicillin (100 U´mL21) and streptomycin(100 mg´mL21), in a humid atmosphere containing 5% CO2

at 37 8C. Cells grown to 90% confluency were exposed at 37 8Cto serial tenfold dilutions of the purified PLCs (214 mg´mL21)in 200 mL of medium per well and cell viability was measured24 h later using a neutral red assay [26]. All experiments wereperformed three times, with four replicate samples each time.

Myotoxicity was estimated by measurement of the creatinekinase (CK) release to the plasma after injection of the PLCs inthe right gastrocnemius muscle in groups of four Swiss-Webstermice, as described [27]. Animals were housed, fed and handledaccording to the principles and practices approved by Vicer-rectoria de InvestigacioÂn, Universidad de Costa Rica. The CKactivity was determined using either a colorimetric or a kineticassay (Sigma, kits 520 or CK-10). Plasma CK activity wasmaximal at 3 h after injection of the wild-type a-toxin andtherefore subsequent measurements were performed at thistime. A linear dose±response curve was obtained using0.5±2.5 mg of wild-type a-toxin, and therefore 1.5 mg of eachvariant was used. Mice were sacrificed by cervical dislocation24 h after injection and the injected gastrocnemius muscle wasdissected and immersed in Karnovsky's fixative solution (2.5%glutaraldehyde, 2% paraformaldehyde, 0.1 m phosphate buffer,pH 7.4) for 2 h. The tissue was postfixated with osmiumtetroxide, dehydrated and embedded in Spurr resin. Sections of1 mm were prepared, stained with toluidine blue and observedunder the light microscope.

For intravital microscopy experiments male mice (18±22 g)were anesthetized with an intraperitoneal injection of sodiumpentobarbital (2 mg per 100 g body weight), placed on a water-heated bed (at 37 8C) and the cremaster muscle exposed [28].Solutions of 1.5 mg of PLCs in 20 mL of NaCl/Pi were appliedto the exposed muscle, which was immediately covered with6-mm Mylarw sheet to prevent dehydration. Control experi-ments were performed in muscles treated with 20 mL of

5192 A. Alape-GiroÂn et al. (Eur. J. Biochem. 267) q FEBS 2000

NaCl/Pi alone. The alterations induced by the toxins wererecorded during 25 min.

R E S U LT S

Choice of substitutions and molecular modelling of a-toxinvariants

The residues of a-toxin selected for mutagenesis were Asp269,Asp336, Tyr275, Tyr307, and Tyr331, which are located in theputative membrane-binding region of the C-terminal domain,except Tyr307, which is part of the interface between theN- and C-terminal domains (Fig. 1). The selection was basedon structural comparisons of the C-terminal domain of a-toxinwith that of the weakly toxic C. bifermentans PLC and C2domains from eukaryotic proteins [9]. Tyr367, which is distantfrom the putative membrane-binding surface and the N andC-terminal domain interface (Fig. 1), was mutated as a control.The introduced residues were chosen on the basis of theirpropensity to adopt the same secondary structure as those ofthe corresponding wild-type residues [29]. Asp residues weresubstituted by Asn to evaluate the role of the carboxylategroups, whereas Tyr residues were replaced by Phe todetermine the role of the hydroxy groups. In addition, Tyr275and Tyr331 were substituted by Asn and Leu, respectively, tomimic locally the structure of the C. bifermentans PLC. Themutant a-toxin variants were recognized identically as the wild-type toxin by the mAbs 3A4D10, 3A4F2, DY2F5, 9F3A6,A5A11 in an ELISA (not shown), demonstrating that despitesubstitutions several epitopes remain intact.

The hemolytic and cytotoxic activities are reduced byaspartic acid and tyrosine substitutions

The variants D269N and D336N had less than 19 and 11% ofthe hemolytic activity, respectively (Fig. 2A) and displayed acytotoxic potency 103-fold lower than that of the wild-typea-toxin (Fig. 2B). The variant Y275N had only 11% of thehemolytic activity (Fig. 3A) and displayed a cytotoxic potency105-fold lower than that of the wild-type a-toxin (Fig. 3B),

Fig. 1. Schematic representation of the

C. perfringens a-toxin structure. The figure is

shaded from red for residues farthest from the

membrane to blue for residues likely to be

inserted into the membrane. The plane of the

membrane is determined by the three calcium

ions in the C-terminal domain (which are

thought to coordinate the phospholipid

headgroups of the target membrane). Mutated

residues are shown in green ball-and-stick. A

phospholipid molecule has been modeled in the

active site in transparent ball-and-stick.

Fig. 2. Hemolytic and cytotoxic activities of the wild-type (Wt)

C. perfringens a-toxin and variants with single substitutions in aspartic

acid residues. (A) Hemolytic activity. Sheep erythrocytes in agarose gels

were exposed for 24 h at 37 8C to 0.35 mg of each a-toxin variant and the

diameter of hemolysis was measured. The results are expressed as the

percentage of hemolytic activity, considering 100% to be the diameter of

the hemolytic halo induced by the wild-type a-toxin. (B) Cytotoxic activity.

Cells were exposed for 24 h at 37 8C to serial tenfold dilutions of each

a-toxin variant: wild-type toxin (O); D336N (�); D269N (W). Cell viability

was determined using the neutral red assay as described in Materials and

methods. The results are expressed as the percentage of neutral red uptake

in relation to controls incubated without toxin.

q FEBS 2000 Residues critical for toxicity in phospholipase C (Eur. J. Biochem. 267) 5193

showing that the phenolic ring of Tyr275 is crucial for thehemolytic and cytotoxic activities. Furthermore, the variantY275F had 73% of the hemolytic activity (Fig. 3A) andexhibited a cytotoxic potency 102-fold lower than the wild-typea-toxin (Fig. 3B), demonstrating the importance of the hydroxygroup at 275 for those toxic activities. Similarly, the variantsY331F and Y331L had 36 and 30% of the hemolytic activity,respectively (Fig. 3A), and displayed a cytotoxic potency 103-to 104-fold lower than the wild-type toxin (Fig. 3C), showingthe importance of the hydroxy group at 331 for toxicity. The

Fig. 3. Hemolytic and cytotoxic activities of the wild-type (Wt)

C. perfringens a-toxin and variants with single substitutions in tyrosine

residues. (A) Hemolytic activity. Sheep erythrocytes in agarose gels

were exposed for 24 h at 37 8C to 0.35 mg of each a-toxin variant and

the diameter of hemolysis was measured. The results are expressed as

the percentage of the hemolytic activity, considering 100% the diameter

of the hemolytic halo induced by the wild-type a-toxin. (B±D) Cyto-

toxic activity. Cells were exposed for 24 h at 37 8C to serial tenfold

dilutions of each a-toxin variant: wild-type toxin (O). (B) Y275F (�);

Y275N (W). (C) Y331F (�); Y331L (W). (D) Y307F (W). Cell viability was

determined using the neutral red assay as described in Materials and

methods. The results are expressed as the percentage of neutral red uptake

in relation to controls incubated without toxin.

Fig. 4. Myotoxic activity of the C. bifermentans PLC, the wild-type

(Wt) C. perfringens a-toxin and genetically engineered a-toxin

variants. Groups of four Swiss-Webster mice were injected with 1.5 mg

of each variant in the right gastrocnemius muscle and plasma CK activity

was measured 3 h later as described in Materials and methods. In panel B,

the results are expressed as percentage of the plasma CK increment,

considering 100% as the CK activity of mice injected with the wild-type

a-toxin.

5194 A. Alape-GiroÂn et al. (Eur. J. Biochem. 267) q FEBS 2000

variant Y307F had about 38% of the hemolytic activity(Fig. 3A), and displayed a cytotoxic potency 103-fold lowerthan that of the wild-type a-toxin (Fig. 3D), whereas thecontrol variant Y367F had identical hemolytic and cytotoxicactivities as the wild-type a-toxin (not shown).

The myotoxicity of a-toxin requires the sphingomyelinaseactivity and the C-terminal domain

Although it is known that a-toxin causes extensive myonecrosis[30,31], there are no previous studies which report therelationship between the structure of the protein and itsmyotoxic activity. The wild-type recombinant a-toxin injectedintramuscularly induced in 3 h a 35-fold increase in plasmalevels of CK activity as compared to control mice injected withNaCl/Pi (Fig. 4A). In contrast, the C. bifermentans PLC onlyproduced less than a twofold increase in the plasma CK level(Fig. 4A). Microscopic assessment of muscle damage in tissuesamples obtained 24 h after toxin injection showed goodcorrelation with increments in plasma CK activity (not shown).Asp56 is a Zn21 ligand in the active site (Fig. 1) and itsreplacement does not affect the membrane-binding capacity ofthe holotoxin [32], although it completely abrogates the leci-thinase and sphingomyelinase activities [18,32]. Intramuscularinjection of the enzymatically inactive variant D56N [18,32]failed to induce a significant myotoxic effect, and only caused aslight increase in CK plasma activity (Fig. 4B), indicating thatthe catalytic capacity of a-toxin is required for its myotoxicity.A truncated a-toxin containing only the 249 N-terminalresidues retains the lecithinase activity but lacks the sphingo-myelinase activity [12]. This fragment induced a 50-fold lowerincrease in CK plasma activity than the holotoxin (Fig. 4B).Thus, the data demonstrate that the lecithinase activity per seis not sufficient to confer myotoxicity and shows that theC-terminal domain is required for full myotoxic activity.

The effects of a-toxin on the mouse cremaster muscle and itsmicrocirculation were studied by intravital microscopy. Hyper-contraction and conspicuous morphological alterations of themuscle fibers, likely due to the membrane-damaging effect ofa-toxin, were already evident 1±2 min after toxin application.

The toxin also caused a progressive reduction in the blood flow.About 2±5 min after toxin exposure thrombi began to form,first in the walls of the venules and then in arterioles, giving offnumerous emboli. These intravascular aggregates grew pro-gressively in size and number, disturbing the local blood flowand finally leading to a complete halt of the circulation within a15±20-min period. In contrast, none of the described eventswere observed in preparations treated with the enzymaticallyinactive variant D56N or the truncated a-toxin.

The myotoxic activity is reduced by substitutions in residuesof the putative membrane-interacting surface

The variants D269N, D336N, Y275N and Y331L were selectedfor studies of myotoxicity in vivo because they had the lowesthemolytic and cytotoxic activities among the variants used inthis study. These variants showed less than 12% of the myo-toxic potency of the wild-type toxin, as judged by the increasein the CK plasma activity observed after intramuscularinjection (Fig. 5). Furthermore, they induced negligible tissuedamage in the mouse cremaster muscle.

D I S C U S S I O N

We have studied in vitro and in vivo the toxic activities ofseveral genetically engineered a-toxin variants whose lecithin-ase activity has been previously characterized [9]. These vari-ants harbor single amino-acid substitutions in residues locatedwithin the proposed membrane-interacting region of theC-terminal domain. As the substituted residues are located atthe surface of the protein and most substitutions were conser-vative, it was expected that they would not interfere withcorrect folding of the protein. Accordingly, all the variants wereidentically recognized by several mAbs as the wild-type toxinand restrained energy minimization did not result in an alter-ation of the conformation of the protein. Furthermore, previouspulse chase experiments have shown that these variants havethe same thermodynamic stability as the wild-type a-toxin andbehave in an identical way in gel filtration, ion-exchangechromatography and native gel electrophoresis [9]. Thus,although a detailed analysis of the different side-chain±side-chain interactions that arise in each variant must await theirthree-dimensional structure determination, several lines ofevidence indicate that all of them have a similar conformationto the wild-type a-toxin.

The results obtained with the variants D269N, D336N andY307F showed that Asp269, Asp336 and Tyr307 are crucial toconfer full hemolytic and cytotoxic activities to a-toxin, inagreement with the hypothesis proposed by Naylor et al. [7].Asp269 and Asp336 are Ca21 ligands in a-toxin [11] and thecorresponding residues are also known to bind Ca21 in severalC2 domains of eukaryotic proteins [7,9]. Tyr307 makes ahydrogen bond to Asp298, which is also a Ca21 ligand [11] andtherefore could also be required for a proper interaction ofa-toxin with the target membrane. In addition, Tyr307 ispart of the interface between the C-terminal domain and theN-terminal domain, and so could be involved in signalingbetween the major membrane binding surface and the activesite. Thus, the lowered hemolytic and cytotoxic activities of thevariants D269N, D336N and Y307F are likely explained by areduced capacity to interact with the membrane of the targetcells (and, in the case of Y307F, also by a reduced capacity tocommunicate such interaction to the active site).

Although substitution of Tyr275 or Tyr331 with Phe reducedthe capacity of a-toxin to hydrolyse aggregated PtdCho only by

Fig. 5. Myotoxic activity of the wild-type (Wt) C. perfringens a-toxin

and variants with single substitutions in residues of the C-terminal

domain. Groups of four Swiss-Webster mice were injected with 1.5 mg of

each variant in the right gastrocnemius muscle and plasma CK activity was

measured 3 h later as described in Materials and methods. The results are

expressed as percentage of the plasma CK increment, considering 100% as

the CK activity of mice injected with the wild-type a-toxin.

q FEBS 2000 Residues critical for toxicity in phospholipase C (Eur. J. Biochem. 267) 5195

20 and 5%, respectively [9], these substitutions affected thehemolytic and cytotoxic activities to a larger extent. Further-more, the Y275N and Y331L substitutions also caused adramatic reduction in the hemolytic, cytotoxic and myotoxicactivities of a-toxin, underscoring the importance of Tyr275and Tyr331 for toxicity. Since these tyrosine residues are Asnand Ile, respectively, in the C. bifermentans PLC, these substi-tutions could explain, at least partially, the lower hemolytic,cytotoxic and myotoxic activities of this enzyme in comparisonwith a-toxin.

C. perfringens a-toxin induces the upregulation of adhesionmolecules in leukocytes and endothelial cells as well as theproduction of platelet activating factor [33,34]. Thus, throm-botic events are thought to reduce the blood supply in tissuesinfected with C. perfringens, promoting the anaerobic environ-ment required for bacterial growth [31,35]. In this work, weobserved in vivo that a-toxin, besides damaging the musclefibers, also induces the formation of intravascular aggregateswhich dramatically affect the microcirculation of the mousecremaster muscle. The myotoxicity as well as the capacity toinduce thrombotic events is impaired in the mutant variantsD269N, D336N, Y275N and Y331L. These results give astarting point for further in vivo studies aimed at developinggenetically engineered vaccines for gas gangrene, a disease ofincreasing significance in the elderly and diabetics, especiallyin those who have undergone lower limb surgery [36,37].

The C-terminal domain of a-toxin resembles C2 domains ofeukaryotic proteins, which despite a high sequence diversity arestructurally similar [8]. The binding of Ca21 to C2 domainscommonly promotes a reversible binding to membranes andmight cause conformational changes which lead to membranepenetration or trigger local changes in surface potential, henceinducing a selective interaction with proteins of the targetmembrane [38,39]. Electrostatic and hydrophobic interactionsmake variable contributions to membrane binding depending onthe residues flanking the Ca21 ligands in the loops [40]. Thus,the side-chains of residues surrounding the Ca21 binding sitesdefine the specificity of each C2 domain for different phos-pholipid types. In the case of a-toxin, it binds and disruptsartificial membranes with PtdCho or SM but not those withPtdSer, PtdEtn or PtdGro [41]. However, our understanding ofthe molecular interactions involved in the binding of a-toxin tomembranes is incomplete, particularly regarding the relativecontributions of electrostatic and hydrophobic interactions. Itis not known whether Ca21 ions only anchor a-toxin via abridging mechanism, or whether they trigger membrane pene-tration, and/or protein±protein interactions. Tyrosine and tryp-tophan residues have been suggested to play a critical role inthe interaction of peptides with the interfacial region of lipidbilayers [42]. Indeed, several lines of evidence indicate thattyrosine and tryptophan residues, because of their amphipathicside chains, are ideally suited to reside in the complexelectrostatic environment of the bilayer interface: (a) phenolicand indolic ring analogues seek out a shallower locations thanaromatic rings, and preferentially partition at the boundarybetween the acyl chain and the polar region of phospholipidbilayers [43]; (b) membrane proteins are enriched in tyrosineand tryptophan residues near the headgroup of the lipid bilayer[44]; (c) the capacity of the membrane-interacting oligopeptidesgramicidins to modulate the bilayer structure (inducing hexagonalphase formation) depends on the presence of tryptophan residues,which can be replaced with tyrosine, but not with phenylalanine[45]; (d) tyrosine and tryptophan residues are critical for theanchoring of pancreatic colipase, and cytosolic and secretedphospholipases A2 to micellar lipids [46±48]. Further structural

studies of the a-toxin variants with substitutions in Tyr275 andTyr331 will help to elucidate the exact role of these residues inthe interaction of a-toxin with phospholipid bilayers.

A C K N O W L E D G M E N T S

We thank J. NunÄez and J. Sanabria (Instituto Clodomiro Picado) for

valuable help, as well as Prof. H. JoÈrnvall and Dr I. Florin (Karolinska

Institutet) for critical reading of the manuscript. This work was supported

by grants from the Swedish Cancer Society (3826-B96-01XAB), Consejo

Nacional de Investigaciones CientõÂficas y TecnoloÂgicas de Costa Rica

(CONICIT), VicerrectorõÂa de InvestigacioÂn, Universidad de Costa Rica

(741-98-287), the Swedish Medical Research Council (16X-05969) and the

Karolinska Institutet Research Funds.

R E F E R E N C E S

1. Songer, J.G. (1997) Bacterial phospholipases and their role in

virulence. Trends Microbiol. 5, 156±161.

2. Titball, R.W. (1993) Bacterial phospholipases C. Microbiol. Rev. 57,

347±366.

3. Titball, R.W. (1997) Clostridial phospholipases. In The Clostridia:

Molecular Biology and Pathogenesis. (Rood, J.I., McClane, B.A.,

Songer, J.G. & Titball, R.W., eds.), pp. 223±242. Academic Press,

New York.

4. Ninomiya, M., Matsushita, O., Minami, J., Sakamoto, H., Nakano, M.

& Okabe, A. (1994) Role of alpha-toxin in Clostridium perfringens

infection determined using recombinants of C. perfringens and

Bacillus subtilis. Infect. Immun. 62, 5032±5039.

5. Awad, M.M., Bryant, A.E., Stevens, D.L. & Rood, J. (1995) Virulence

studies on chromosomal a-toxin and é-toxin mutants constructed by

allelic exchange provide genetic evidence for the essential role

of a-toxin in Clostridium perfringens-mediated gas gangrene.

Mol. Microbiol. 15, 191±202.

6. Titball, R.W., Naylor, C.E. & Basak, A.K. (1999) The Clostridium

perfringens a-toxin. Anaerobe 5, 51±64.

7. Naylor, C.E., Eaton, J.T., Howells, A., Justin, N., Moss, D.S., Titball,

R.W. & Basak, A.K. (1998) Structure of the key toxin in gas

gangrene. Nature Struct. Biol. 5, 738±746.

8. Nalefski, E.A. & Falke, J.J. (1996) The C2 domain calcium-binding

motif: structural and functional diversity. Protein Sci. 5, 2375±2390.

9. Guillouard, I., Alzari, P.M., Saliou, B. & Cole, S. (1997) The carboxy-

terminal C2-like domain of the a-toxin from Clostridium perfringens

mediates calcium-dependent membrane recognition. Mol. Microb. 26,

867±876.

10. Nagahama, M., Michiue, K., Mukai, M., Ochi, S. & Sakurai, J. (1998)

Mechanism of membrane damage by Clostridium perfringens alpha-

toxin. Microbiol. Immunol. 42, 533±538.

11. Naylor, C.E., Jepson, M., Crane, D.T., Titball, R.W., Miller, J., Basak,

A.K. & Bolgiano, B. (1999) Characterization of the Calcium-binding

C-terminal domain of Clostridium perfringens alpha-toxin. J. Mol.

Biol. 294, 757±770.

12. Titball, R.W., Leslie, D.L., Harvey, S. & Kelly, D. (1991) Hemolytic

and sphingomyelinase activities of Clostridium perfringens alpha-

toxin are dependent on a domain homologous to that of an enzyme

from the human arachidonic acid pathway. Infect. Immun. 59,

1872±1874.

13. Flores-DõÂaz, M., Alape-GiroÂn, A., Titball, R.W., Moos, M., Guillouard,

I., Cole, S., von Eichel-Streiber, C., Florin, I. & Thelestam, M. (1998)

UDP-glucose deficiency causes hypersensitivity to the cytotoxic

effect of Clostridium perfringens phospholipase C. J. Biol. Chem.

273, 24433±24438.

14. Jepson, M., Howells, A., Bullifent, H.L., Bolgiano, B., Crane, D.,

Miller, J., Holley, J., Jayasekera, P. & Titball, R.W. (1999) Differ-

ences in the carboxy-terminal (putative phospholipid binding)

domains of Clostridium perfringens and Clostridium bifermentans

phospholipases C influence the hemolytic and lethal properties of

these enzymes. Infect. Immun. 67, 3297±3301.

15. Sakurai, J., Fujii, Y., Torii, K. & Kobayashi, K. (1989) Dissociation of

5196 A. Alape-GiroÂn et al. (Eur. J. Biochem. 267) q FEBS 2000

various biological activities of Clostridium perfringens alpha toxin by

chemical modification. Toxicon 27, 317±323.

16. Titball, R.W., Hunter, S.E.C., Martin, K.L., Morris, B.C., Shuttleworth,

A.D., Rubidge, T., Anderson, D.W. & Kelly, D.C. (1989) Molecular

cloning and nucleotide sequence of the alpha-toxin (phospho-

lipase C) of Clostridium perfringens. Infect. Immun. 57, 367±376.

17. Saint-Joanis, B., Garnier, T. & Cole, S.T. (1989) Gene cloning shows

the alpha-toxin of Clostridium perfringens to contain both sphingo-

myelinase and lecithinase activities. Mol. Gen. Genet. 219, 453±460.

18. Guillouard, I., Garnier, T. & Cole, S.T. (1996) Use of site-directed

mutagenesis to probe structure-function relationships of alpha-toxin

from Clostridium perfringens. Infect. Immun. 64, 2440±2444.

19. Sato, H., Chiba, J. & Sato, Y. (1989) Monoclonal antibodies against

alpha toxin of Clostridium perfringens. FEMS Microbiol. Lett. 59,

173±176.

20. Logan, A.J., Williamson, E.D., Titball, R.W., Percival, D.A., Shuttle-

worth, A.D., Conlan, J.W. & Kelly, D.C. (1991) Epitope mapping

of the alpha-toxin of Clostridium perfringens. Infect. Immun. 59,

4338±4342.

21. Jones, A., Zou, J.-y. & Cowna, S.W. (1991) Improved methods for

binding protein models in electron density maps and the location of

errors in these models. Acta Crystallogr. A 47, 110±119.

22. BruÈnger, A.T. (1992) X-PLOR Manual, Version 3.1. A System for X-Ray

Crystallography and NMR. Yale University Press, New Haven, CT.

23. Esnouf, R. (1997) An extensively modified version of MOLSCRIPT

that includes greatly modified colouring capabilities. J. Mol. Graphics

15, 132±134.

24. Merrit, E.A. & Murphy, M.E.P. (1994) RASTER 3D, Version 2.0 ± a

program for photorealistic molecular graphics. Acta Crystallogr. D

50, 869±873.

25. Nicholls, A., Sharp, K.A. & Honig, B. (1991) Protein folding and

association: insights from the interfacial and thermodynamic proper-

ties of hydrocarbons. Proteins Struct. Funct. Genet. 11, 281±296.

26. Flores-Diaz, M., Alape-GiroÂn, A., Persson, B., Pollesello, P., Moos, M.,

von Eichel-Streiber, C., Thelestam, M. & Florin, I. (1997) UDP-Glc

deficiency caused by a single point mutation in the UDP-Glc

pyrophosphorylase gene. J. Biol. Chem. 272, 23784±23791.

27. GutieÂrrez, J.M., Arroyo, O. & BolanÄos, R. (1980) Mionecrosis,

hemorragia y edema inducidos por el veneno de Bothrops asper en

ratoÂn blanco. Toxicon 18, 603±610.

28. Lomonte, B., Lundgren, J., Johanson, B. & Bagge, U. (1994) The

dynamics of local tissue damage induced by Bothrops asper snake

venom and myotoxin II on the cremaster muscle: an intravital and

electron microscopic study. Toxicon 32, 41±55.

29. Bordo, D. & Argos, P. (1991) Suggestions for `̀ safe'' residue

substitutions in site-directed mutagenesis. J. Mol Biol. 217, 721±729.

30. Strunk, S.W., Chauncey, W.S. & Blumberg, J.M. (1967) Ultrastructural

studies on the lesion produced in skeletal muscle fibers by crude

type A Clostridium perfringens toxin and its purified alpha fraction.

Am. J. Pathol. 50, 89±107.

31. Stevens, D.L., Tweten, R.K., Awad, M.M., Rood, J.I. & Bryant, A.E.

(1997) Clostridial gas gangrene: evidence that a and é toxins

differentially modulate the immune response and induce acute tissue

necrosis. J. Infect. Dis. 176, 189±195.

32. Nagahama, M., Nakayama, T., Michiue, K. & Sakurai, J. (1997)

Site-specific mutagenesis of Clostridium perfringens alpha-toxin:

replacement of Asp-56, Asp-130 or Glu-152 causes loss of enzymatic

and hemolytic activities. Infect. Immun. 65, 3849±3492.

33. Bryant, A.E. & Stevens, D.L. (1996) Phospholipase C and perfringo-

lysin O from Clostridium perfringens upregulate endothelial cell-

leukocyte adherence molecule 1 and intracellular leukocyte

adherence molecule 1 expression and induce interleukin-8 synthesis

in cultured human umbilical vein endothelial cells. Infect. Immun. 64,

358±362.

34. Bunting, M., Lorant, D.E., Bryant, A.E., Zimmerman, G.A., McIntyre,

T.M., Stevens, D.L. & Prescot, S.M. (1997) Alpha-toxin from

Clostridium perfringens induces proinflammatory changes in

endothelial cells. J. Clin. Invest. 100, 565±574.

35. Ellemor, D.M., Baird, R.N., Awad, M.M., Boyd, R.L., Rood, J.I. &

Emmins, J.J. (1999) Use of genetically manipulated strains of

Clostridium perfringens reveals that both alpha-toxin and theta-toxin

are required for vascular leukostasis to occur in experimental gas

gangrene. Infect. Immun. 67, 4902±4907.

36. Williamson, E.D. & Titball, R.W. (1993) A genetically engineered

vaccine against the alpha-toxin of Clostridium perfringens protects

mice against experimental gas gangrene. Vaccine 11, 1253±1258.

37. Chapnick, E.K. & Abter, E.I. (1996) Necrotizing soft-tissue infections.

Infect. Dis. Clin. North. Am. 10, 835±855.

38. Rizo, J. & SuÈdhof, T.C. (1998) C2-domains, structure and function of a

universal Ca21-binding domain. J. Biol. Chem. 273, 15879±15882.

39. Wang, T., Pentyala, S., Elliot, J.T., Dowal, L., Gupta, E., Rebecchi, M.J.

& Scarlata, S. (1998) Selective interaction of the C2 domains of

phospholipase C-b1 and -b2 with activated Gaq subunits: an

alternative function for C2-signaling modules. Proc. Natl Acad. Sci.

USA 96, 7843±7846.

40. Davletov, B., Persic, O. & Williams, R.L. (1998) Calcium-dependent

membrane penetration is a hallmark of the C2 domain of cytosolic

phospholipase A2 whereas the C2A domain of synaptotagmin binds

membranes electrostatically. J. Biol. Chem. 273, 19093±19096.

41. Nagahama, M., Michiue, K. & Sakurai, J. (1996) Membrane-damaging

action of Clostridium perfringens alpha-toxin on phospholipid

liposomes. Biochim. Biophys. Acta 1280, 120±126.

42. Schiffer, M., Chang, C.H. & Stevens, F.J. (1992) The functions of

tryptophan residues in membrane proteins. Protein Eng. 5, 213±214.

43. Kachel, K., Asuncion-Punzalan, E. & London, E. (1995) Anchoring of

tryptophan and tyrosine analogs at the hydrocarbon±polar boundary

in model membrane vesicle: parallax analysis of fluorescence

quenching induced by nitroxide-labeled phospholipids. Biochemistry

34, 15475±15479.

44. Reithmeier, R.A.F. (1995) Characterization and modelling of mem-

brane proteins using sequence analysis. Curr. Opin. Struct. Biol. 5,

491±500.

45. Killian, J.A. (1992) Gramicidin and gramicidin±lipid interactions.

Biochem. Biophys. Acta 1113, 391±425.

46. Van den Berg, B., Tessari, M., Bolens, R., Dijkman, R., Kaptein, R.,

de haas, G.H. & Verheij, H.M. (1995) Solution structure of porcine

pancreatic phospholipase A2 complexed with micelles and a

competitive inhibitor. J. Biomol. NMR 5, 110±121.

47. Hermoso, J., Pignol, D., Penel, S., Roth, M., Chapus, C. & Fontecilla-

Camps, J.C. (1997) Neutron crystallographic evidence of lipase±

colipase complex activation by a micelle. EMBO J. 16, 5531±5536.

48. Glomset, J.A. (1999) Protein±lipid interactions on the surfaces of cell

membranes. Curr. Opin. Struct. Biol. 9, 425±427.

q FEBS 2000 Residues critical for toxicity in phospholipase C (Eur. J. Biochem. 267) 5197