alpha-toxin of staphylococcus aureuss. aureus alpha-toxin 735 at sedimentation equilibrium was...

MICROBIOLOGICAL REVIEWS, Dec. 1991, p. 733-7510146-0749/91/040733-19$02.00/0Copyright X 1991, American Society for Microbiology

Vol. 55, No. 4

Alpha-Toxin of Staphylococcus aureusSUCHARIT BHAKDIl* AND J0RGEN TRANUM-JENSEN2

Institute of Medical Microbiology, University of Mainz, Augustusplatz, W-6500 Mainz, Federal Republic of Germany,'and Anatomy Institute C, University of Copenhagen, DK 2200 Copenhagen N, Denmark2

DISCOVERY AND INCIDENCE ..................................................................... 733EARLY FUNCTIONAL STUDIES.................................................................... 734IDENTIFICATION OF ALPHA-TOXIN AS A PORE FORMER ....................................................734STRUCTURE AND PROPERTIES OF ALPHA-TOXIN ...............................................................735MODES OF PRIMARY BINDING TO TARGET MEMBRANES ...................................................737MODE OF PORE FORMATION .................................................................... 738STRUCTURE AND COMPOSITION OF PORES .................................................................... 738FUNCTIONAL PROPERTIES OF TOXIN PORES: DAMAGE TO LIPOSOMES AND

HEMOLYTIC CAPACITY .................................................................... 740STRUCTURE-FUNCTION RELATIONSHIPS .................................................................... 740EFFECTS OF PORE FORMATION ON TARGET CELLS ...........................................................741

Cell Death .................................................................... 741Secondary Effects......................................................................741

(i) Stimulation of eicosanoid production .................................................................... 742(ii) Secretory processes .................................................................... 742(iii) Cellular contractile dysfunction .................................................................... 743(iv) Activation of endonucleases .................................................................... 743(v) Release of cytokines .................................................................... 743(vi) Membrane alterations .................................................................... 743(vii) Other effects .................................................................... 743(viii) Repair processes .................................................................... 744

Alpha-Toxin as a Biological Tool .................................................................... 744BIOLOGICAL RELEVANCE OF ALPHA-TOXIN .................................................................... 745

Relevance in Animal Models.................................................................... 746Relevance in Humans.................................................................... 746

CONCLUSIONS AND PERSPECTIVES .................................................................... 747ACKNOWLEDGMENTS ..................................................................... 747REFERENCES .................................................................... 747

DISCOVERY AND INCIDENCE

Micrococci were first described by Koch in 1878 (97), andthe identification of Staphylococcus aureus as a humanpathogen followed shortly thereafter through the work ofOgston (117). In the following decades, it became apparentthat S. aureus ranks among the most common causes ofbacterial infections in humans, producing a broad spectrumof diseases ranging from superficial skin suppurations tolife-threatening septicemias. Along with Escherichia coli, italso heads the list of agents that are responsible for hospital-acquired infections.Soon after their discovery, S. aureus isolates were ob-

served to elaborate soluble substances that evoked inflam-matory reactions after inoculation into experimental animals(51, 161a). The production of one or several hemolyticagents could be detected through cultivation on blood agar,and development of a clear zone of beta-hemolysis was oftentaken as a criterion for diagnosis of this bacterium. Thesubsequent introduction of a positive coagulase test as thesole criterion for identifying S. aureus did not generate anyconflict, since beta-hemolytic strains were, virtually withoutexception, coagulase producers.The first serious research into alpha-toxin was sparked by

* Corresponding author.

a tragedy in the Australian town of Bundaberg in 1928.Twenty-one children who were vaccinated with a diphtheriatoxoid became severely ill, and 12 died. In the course of hisinvestigation into the cause of this disaster, Burnet isolatedS. aureus from the toxoid preparation and reported on thetoxic properties of a cell-free bacterial culture filtrate (37,38). Work performed in the following decades led to theidentification of alpha-toxin as a major cause of the observedtoxicity. The exotoxin is a secreted protein endowed withhemolytic, cytotoxic, dermonecrotic, and lethal properties.It is also responsible for the development of clear zones ofbeta-hemolysis when bacteria are cultured on blood agar.The alpha-toxin gene is located on the chromosome (35,123). By using a gene probe, 20 coagulase-positive clinicalisolates randomly taken from a routine bacteriological labo-ratory were recently found to harbor the alpha-toxin gene(121). In contrast, the gene was not detected in 30 coagulase-negative isolates. Many S. aureus isolates produce onlysmall amounts of alpha-toxin, and its total absence in S.aureus strains that produce toxic shock syndrome toxin hasbeen reported (45). In the latter case, this is due to anonsense mutation that generates a stop codon within thegene (120a). Both coagulase-positive and coagulase-negativestaphylococci may produce other hemolysins, designatedbeta-, gamma-, and delta-toxins, that are molecularly dis-tinct from alpha-toxin and whose roles as virulence factors

733

on April 8, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

http://mmbr.asm.org/

734 BHAKDI AND TRANUM-JENSEN

are poorly understood (11, 59, 130, 168). The present reviewwill be confined to a discussion of alpha-toxin.

EARLY FUNCTIONAL STUDIES

Interested readers are referred to earlier reviews (11, 59,83, 108, 130, 168) for summaries of data on the manyfunctional studies that were performed in the period from1940 to the mid-1960s. The review by Wiseman (168) shouldbe consulted for information on the lethal and dermonecroticdoses determined in various animal models. Although thoseearly studies were performed with material that could notmeet current-day demands for purity, a number of funda-mental observations were made. One remarkable findingrelated to the variation in susceptibility to lysis exhibited byerythrocytes of different animal species (see references 50,83, and 108 for summaries of these data). For example, anapproximately 400-fold-higher concentration of alpha-toxinwas required to lyse human erythrocytes compared withrabbit erythrocytes. In addition to erythrocytes, a widevariety of cells including macrophages, skin epithelium,rabbit kidney cells, fibroblasts, smooth muscle cells, andmast cells responded to toxin attack (see reference 83 forcompilations). Diverse reactions in these cells were noted;they included liberation of lysosomal enzymes and spasticcontraction of smooth muscle cells (83, 171). These datawere significant because they gave an early indication thatthe action of alpha-toxin was not confined to erythrocytesand that nucleated cells might respond to toxin attack with aspectrum of secondary reactions. Although it is possible thatsome of the observed phenomena were evoked by contam-inants in the toxin preparations, this basic message hasproved to be correct.At that time, neither the mechanism of toxin action nor the

cause of its cellular effects was understood in any detail.Following the introduction of analyses of lysis kinetics byBernheimer (for reviews, see references 11 and 108), studiesconducted by Cooper et al. (49) led them to propose thathemolysis involved a defined sequence of events: first, theinteraction of toxin with cells; second, release of K+ ions;and, finally, hemolysis. How the hemolytic action of alpha-toxin could be conceptually linked to its capacity to provokepathological reactions in nucleated cells remained enigmatic,as did the cause of its lethal effect. At one stage, lethalitywas proposed to derive from attack on the central nervoussystem, since electroencephalography indicated collapse ofbrain bioelectric activity (52, 83). In support of this conten-tion, one study purported to show that intravenously appliedradiolabeled toxin accumulated not only in the kidneys andlungs but also in the brain (85). This claim has been con-tested in a recent investigation in which functionally intactradiolabeled toxin was not detected in the central nervoussystem of mice that received lethal and sublethal toxin doses(28). The question of whether toxin exerts pathophysiologi-cally relevant effects via its action on the central nervoussystem should therefore today be regarded critically. Otherearlier work provided evidence that alpha-toxin promotedvasospasm of blood vessels (146, 147), possibly as a result ofits action on smooth muscle cells (42, 171). According to apreliminary report (mentioned in reference 83), cardiacoutput also declined after toxin application, supporting thenotion that lethality may be due primarily to cardiovascularcollapse. As will be discussed in this review, evidence hasnow accumulated that alpha-toxin preferentially attacks en-dothelial cells and thrombocytes and that secondary reac-tions triggered in these cells can lead to deleterious distur-

bances in the microcirculation and to the development ofpulmonary edema. These recent results therefore supportthe concept that the acute lethal action of alpha-toxin isprobably due to its effects on cells that are involved in themaintenance of hemostasis.

Despite the numerous studies describing biological effectsof alpha-toxin in vivo and in vitro, its relevance in humanshas remained the subject of continuous debate. The intrinsicresistance of human erythrocytes against toxin action fos-tered the impression that a pathogenic role for this agentmight be confined to certain animal species only. Staphylo-cocci secrete a number of other toxic substances includingleukocidin, coagulase, proteases, and lipase, which manyconsider to contribute more significantly than alpha-toxin tobacterial virulence in humans. In this context, however, oneearly study deserves special mention. Siegel and Cohen (136)reported that application of alpha-toxin to platelet-rich hu-man plasma induced platelets to undergo shape change andaggregation, processes that were accompanied by a netprocoagulatory effect. The platelets leaked NAD+ and K+,but not protein; hence, frank cell lysis apparently did notoccur. The authors concluded that the ability of alpha-toxinto directly influence a major component of hemostasis mightbe clinically relevant. They anticipated that they had "iden-tified a human cell (or cell fragment) highly susceptible tothis agent" (136). At approximately the same time, Bern-heimer and Schwartz also noted that rabbit platelets werehighly sensitive to the action of alpha-toxin (14). A confir-matory report subsequently provided electron micrographsthat demonstrated swelling, but no lysis, of toxin-treatedhuman platelets (105).More than 20 years elapsed before this important theme

was further investigated. The observation that human plate-lets are highly sensitive targets for alpha-toxin was con-firmed (23). The need to rectify the prevailing dogma thatalpha-toxin does not damage human cells was underlined bythe subsequent identification of monocytes (22) and endo-thelial cells (140) as two other highly susceptible human celltypes. The procoagulatory effect of alpha-toxin was shownto derive from assembly of prothrombinase complexes onthe platelet surface, a major contributing factor being theexocytotic release of factor V from platelet granules (7).These findings are mentioned here because they introduce usto two basic questions. Why does alpha-toxin damage hu-man platelets but not erythrocytes or granulocytes? Why dodamaged platelets secrete factor V? The resolution of theseproblems has led to a general appreciation of how mem-brane-damaging toxins can ultimately evoke pathophysiolog-ical reactions that may be relevant to the pathogenesis ofdisease.

IDENTIFICATION OF ALPHA-TOXIN AS APORE FORMER

The modern era of alpha-toxin research was heralded bypublications in the mid-1960s describing methods for isolat-ing highly purified toxin (13, 98, 102). Molecular weightsgiven in the early investigations ranged from 26,000 to 36,000(for reviews of the earlier literature, see reference 108).Several groups had certainly obtained very pure alpha-toxin(13, 102, 137, 138), and one molecular weight analysis byultracentrifugation was performed by Arbuthnott et al. in1967 (5). A later detailed study (17) corroborated thesefindings in showing that native alpha-toxin is a hydrophilicmolecule that is present in monomeric form in aqueoussolution. The molecular weight of the monomer determined

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from


S. AUREUS ALPHA-TOXIN 735

at sedimentation equilibrium was 33,000 to 34,000, and thesedimentation coefficient was 3.3S (17). These data were infair agreement with previous results of Lominski et al. (102),McNiven et al. (112), and Six and Harshman (137, 138), whohad reported a sedimentation coefficient of 3S and a molec-ular weight estimate of 28,000 to 30,000. Pioneering work onthe mechanism of toxin action was conducted in 1967 to 1973by Arbuthnott and coworkers and Bernheimer and cowork-ers (5, 6, 12, 60, 61, 165). These investigators reported thatmembranes of cells treated with high doses of alpha-toxincarried annular structures of approximately 10 nm in diam-eter that could be detected by electron microscopy. Fortu-itously, similar structures were found when purified toxinsolutions were heated to 60°C (in the absence of cells).Molecular weight approximations by Lominski et al. (102),Arbuthnott et al. (5), Forlani et al. (57), and McNiven et al.(112) finally led to the proposition that the rings represented12S hexamers formed through aggregation of native 3S toxinmonomers (6, 11, 59-61). Similar ring structures could alsobe generated on liposomal membranes devoid of protein (6,60). From these observations, the following conclusionswere drawn: (i) alpha-toxin bound firmly to target mem-branes; (ii) binding did not require the presence of proteinreceptors; (iii) binding was accompanied by the formation ofring-structured toxin oligomers, probably hexamers; and (iv)binding presumably caused membrane damage, albeit via anunknown mechanism.

Cassidy et al. (42) confirmed that highly purified toxininteracted with protein-free liposomal membranes and ex-tended these observations by demonstrating that the lipo-somes became damaged and leaked trapped marker mole-cules. Their studies corroborated earlier work by Colaciccoand Buckelew (46), who had demonstrated a disruption oflipid monolayers by alpha-toxin. In 1973 to 1975, Thelestamet al. (150, 151, 153) showed that nucleated cells attacked byalpha-toxin leaked low-molecular-weight markers while re-taining RNA. They consequently alluded to functional"holes" that were apparently produced by alpha-toxin in theplasma membrane. These studies were in accord with theearly observation of Siegel and Cohen (136), since theyindicated that the membrane lesions produced by alpha-toxin were of finite size.Today, it is obvious that the above observations uniformly

pointed to a pore-forming action of alpha-toxin. In themid-1970s an analogous situation had arisen in the comple-ment field. Cells damaged by complement appeared to carryfunctional membrane lesions of finite size, and annularstructures could be observed attached to these membranes.In 1972 Mayer (106) proposed that complement proteins C5to C9 formed the annular structures through their assemblyinto macromolecular CSb-9 complexes that inserted into thebilayer to create aqueous transmembrane pores. The origi-nality of this hypothesis, which was initially greeted withskepticism, lay in the assumption that primarily water-soluble plasma proteins might transform themselves intoamphiphilic moieties, presumably through conformationalchanges leading to an exposure of lipid-binding domains.Mayer's hypothesis was subsequently confirmed by ultra-structural, biochemical, and functional studies, the CSb-9complex thus becoming the first macromolecular pore-form-ing cytolysin to be identified (15, 156).The conceptual basis for the complement work was de-

rived in part from discoveries made in the early 1970s thatsmall amphiphilic molecules of microbial origin such as thepolyene antibiotics as well as peptides such as gramicidin Acould spontaneously insert into cell membranes and form

discrete pores (74, 96). The experience gained in the com-plement field was subsequently applied to alpha-toxin re-search. Toxin-treated erythrocyte membranes were solubi-lized with mild detergents which preserved the porestructures and allowed their isolation and identification asamphiphilic, ring-shaped hexamers carrying lipid- and deter-gent-binding domains (62). Functional studies indicated thatdamaged erythrocyte membranes leaked molecules whoseeffective diameter did not exceed 1 to 2 nm; these dimen-sions corresponded approximately to the size of the channelthat appeared to traverse the interior of the hexamer asobserved by electron microscopy. The concept of transmem-brane pore formation by alpha-toxin, formulated in 1981(62), stated that native toxin is secreted as a hydrophilicmolecule of Mr 34,000. After binding to target membranes,toxin molecules oligomerize to form noncovalently associ-ated, stable hexameric protein complexes. This process isassociated with an exposure of lipid-binding domains, pre-sumably resulting from conformational changes (e.g., un-folding), that enable the hexamer to spontaneously insertinto the membrane. The hollow interior of the hexamershould generate a hydrophilic channel across the lipid bi-layer, and pore formation was proposed to represent theprimary mechanism of membrane damage by this cytolysin(Fig. 1). The pore concept has been widely confirmed in thepast decade (10, 33, 58, 70, 79-81, 114, 149, 155). Followingthe identification of alpha-toxin as the first pore-formingbacterial cytolysin, many other proteinaceous toxins havebeen found to damage membranes in an analogous fashion.The spectrum of pore-forming cytolysins today encompassesnot only complement and these bacterial toxins, but also themembrane-damaging effector proteins of cytotoxic T lym-phocytes, fungi, and higher parasites (24-26).

STRUCTURE AND PROPERTIES OF ALPHA-TOXIN

The alpha-toxin gene was cloned and sequenced by Kehoeand colleagues (56, 66, 95). The deduced amino acid se-quence yielded a molecular weight of 33,400 for the toxinmonomer (56), in excellent agreement with the physico-chemical data. A signal sequence accounts for a precursordetected by Tweten et al. (157). Lee and Birbeck (100) alsodetected a cell-bound, presumptive precursor molecule afterinhibition of toxin secretion by phenethyl alcohol. Theprotein contains no cystein. It is isoelectric at pH 8.5 to 8.6.

Several earlier reports suggested that alpha-toxin pro-duced by different bacterial strains might exhibit significantmolecular heterogeneity (summarized in references 108 and168). However, this contention has not been substantiated inthe more recent literature, and Southern analyses of DNAfrom 20 S. aureus isolates revealed no heterogeneity inrestriction fragments (121). Furthermore, neither immuno-logical nor DNA hybridization tests have led to detection ofany structurally and genetically related microbial products.There is no homology between alpha-toxin and other poreformers, including complement and lymphocytolysins.

Regulation of expression of the alpha-toxin gene is cur-rently under investigation. The toxin gene is present in asingle copy in the bacterial chromosome (120). A trans-active positive control element (termed agr [accessory generegulator] [128] or exp [expression] [115]) was identified byRecsei et al. (128). This element maps between the purB andilv loci (115) and regulates the production of several exopro-teins including staphylokinase and alpha-, beta-, and delta-hemolysins, as well as serine and metalloproteases (115,

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from



C ,'FIG. 1. (A) Negatively stained fragment of rabbit erythrocyte lysed with S. aureus alpha-toxin. Numerous 10-nm ring-shaped structures

are seen over the membrane (arrows). (B) Isolated toxin hexamers in detergent solution. (C) Lecithin liposomes carrying reincorporated toxinhexamers that are seen as stubs along the edge of the liposomal membrane and as rings over the membrane (arrows). Characteristically,liposomes that escape incorporation of the toxin are impermeable to the stain (asterisk). Sodium silicotungstate negative staining. Scale bars,100 nm. Reproduced with permission from Philosophical Transactions of the Royal Society ofLondon Series B (24).

128). Its discovery explained several earlier observations,e.g., that production of many exoproteins in S. aureus isstrictly synchronized and appears mainly during the postex-ponential or stationary phase (1, 48) and that it was possibleto isolate mutants in which expression of several exoproteinswas simultaneously affected (27, 47, 172). The regulatorychromosomal locus has been cloned (115).The toxin does not harbor extensive stretches of hydro-

phobic amino acids. Analysis of the circular dichroism of thedissolved toxin indicates an abundance of ,B-sheet structure(68%) and very little a-helix (10%) (155). According toTobkes et al. (155) and Ikigae and Nakae (79-81), the overallsecondary structure of oligomeric toxin does not signifi-cantly differ from that of the monomer. This could indicatethat oligomerization does not involve extensive alterations inthe secondary structure of the molecule. In a theoreticalanalysis, Menestrina has concluded that at least 10 am-phiphilic n-sheet structures, each of 3 nm putative length,can be constructed (113). The central portion of the toxinsequence contains an abundance of glycine residues. Thehigh probability of random coil here could indicate that thisregion serves as a hinge for folding between the C and Ndomains. In addition, there is at least one stretch of aminoacids that could produce an amphiphilic a-helix. We havepointed out that pore-forming proteins theoretically need notcontain stretches of hydrophobic amino acids, but are able toproduce pores by the association of amphiphilic ,B-sheets ora-helices (25, 26). One surface will thereby interact withlipid, whereas the other will repel apolar membrane constit-uents, thus creating an aqueous passage. To date, thestructural evidence obtained on all pore formers is consistentwith this concept. It is noteworthy that the most detailedstructural knowledge of a pore-forming protein is from abacterial porin (82). A recent projection map shows threerings of electron density in approximate threefold symmetry.

Each ring is interpreted as arising from 13-sheet structuresoriented perpendicular to the membrane plane. The bacterialporins show a high degree of sequence homology, and havelarge amounts of amphipathic ,B-sheets.

Several pore formers including complement componentC9 (154) and E. coli hemolysin (34, 103) show Ca2e-bindingactivity. With these proteins, pore-forming capacity is lostfollowing prolonged incubation in EDTA. Alpha-toxin doesnot share this property, and no requirement for divalentcations for binding and expression of lytic function has beenrecognized.Two methods currently in use are simple enough to be

recommended for preparative isolation of alpha-toxin. Thefirst involves a one-step purification of a fast protein liquidchromatography column (101). The second involves single-step ion-exchange chromatography of concentrated bacterialculture supernatants over CM-Sephadex (22). The purity ofthe toxin preparations obtained by the latter method exceeds95%, as judged by sodium dodecyl sulfate-polyacrylamidegel electrophoresis (SDS-PAGE) and analytical centrifuga-tion. Toxin preparations can be transferred to 50 mM am-monium acetate and lyophilized. In this form, the protein isstable for years at -20°C and for weeks at room tempera-ture. (Earlier findings of marked instability of;the toxin,discussed in reference 168, were obviously due to thepresence of contaminating proteases.) The Am80 of a 0.5-mg/ml solution is approximately 1.0. The toxin is soluble inwater or any aqueous buffer at pH 5 to, 9.X-ray crytallographic analyses of native or hexamerized

toxin are not yet possible. Although large-scale purificationof native toxin is not a problem, attempts to obtain crystalshave uniformly failed owing to the propensity of the proteinto form amorphous or filamentous aggregates. Isolation ofhomogeneous hexamer preparations is demanding, and thepossibility of contamination with oligomers containing fewer

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from



protomers (e.g., tetramers and pentamers) cannot be ex-cluded. The amphiphilic nature of the hexamers renderstheir handling even more difficult.

MODES OF PRIMARY BINDING TO TARGETMEMBRANES

As alpha-toxin became available in a highly purified form,investigations were launched to analyze the mechanismsresponsible for the different susceptibilities of different celltypes and species. The following findings together consti-tuted a puzzle that challenged the wit of investigators(discussed in detail in reference 108). First, erythrocytesfrom different species exhibit widely different susceptibilitiesto the toxin (108). Second, protein-free liposomes can bedamaged by alpha-toxin (42, 60, 165), and liposomes pre-pared with lipids extracted from rabbit or human erythro-cytes exhibit similar sensitivity (42). Third, the relativeamounts of toxin becoming bound to either susceptible orresistant cells appear to be small. As early as in 1965,Bernheimer and Schwartz (14) reported that toxin-mediatedalterations of platelet function were not accompanied bydetectable consumption of the toxin. Arbuthnott later esti-mated that only approximately 5% of '4C-labeled toxinbound to intact rabbit erythrocytes (108). These observa-tions initially appeared to place some constraints on theconcept that alpha-toxin exerted its cytolytic action via firmbinding to the membranes. (On the other hand, alternativeproposals that alpha-toxin might be an enzyme [169, 170] didnot withstand the test of time.)

Against this background, searches were conducted todetect "receptors" on the highly toxin-sensitive rabbiterythrocytes. Definitive binding studies can, however, beperformed only if a suitable tracer that permits labeling tohigh specific activity without significant loss of biologicalfunction is available. In fact, this demand was not met in anypublished study. The preparation of a suitably labeled toxintracer has been difficult, and several studies purportedlydemonstrating the existence of toxin receptors were thusbased on indirect approaches.

Kato and Naiki (90) found that gangliosides bound andinactivated alpha-toxin and therefore proposed a role forthese molecules as surface receptors. However, binding mayhave been due simply to electrostatic interactions. In fact,Buckelew and Colacicco (36, 46) observed that the rate ofpenetration of alpha-toxin was lowest in lipid monolayerscontaining gangliosides, whereas the highest rates weremeasured in monolayers containing cholesterol without gan-gliosides. By extrapolation, gangliosides may actually pre-vent rather than augment the membrane-damaging action ofalpha-toxin.Maharaj and Fackrell have postulated that the erythrocyte

anion antiporter (band 3, capnophorin) is the toxin receptor(104). Their conclusion was based on indirect evidenceincluding: (i) reduction in sensitivity toward toxin actionexhibited by rabbit erythrocytes following pronase or chy-motrypsin digestion (these enzymes cleave band 3); (ii)reduction in sensitivity of the cells through binding of certainlectins (putative shielding of receptors), as well as afterneuraminidase treatment; and (iii) inactivation of alpha-toxinby purified band 3. The collective data could indicate a rolefor sugar residues present on one or several surface glyco-proteins in binding of alpha-toxin. A protective effect oflectins on hemolysis of rabbit erythrocytes was also reportedby Kato et al. (92). The identity of band 3 as the receptormust, however, be shown by more rigorous criteria. The

number of specific binding sites on rabbit erythrocytes isorders of magnitude smaller than the number of band 3molecules (see below), and the possibility that preparationsof the latter were contaminated with a minor componentcannot be dismissed. The contention that band 3 representsthe toxin receptor has been neither explicitly refuted norconfirmed by others. We have found that washing cellsseveral times does not suffice to remove proteases added tocleave band 3 protein. Residual proteases can then cleavetoxin when it is added to the incubation mix and can therebygenerate artifacts (15). We have also not been able todecrease the sensitivity of rabbit erythrocytes toward alpha-toxin by neuraminidase treatment (15). Harshman and Bon-durant (69) found no differences in susceptibility to alpha-toxin of murine erythroblasts compared with mature cellsand adult erythrocytes. Their results therefore did not sup-port the view that any of the molecules which are synthe-sized late during differentiation of erythrocytes (includingband 3) represent the receptor for alpha-toxin on murinecells. Mouse erythrocytes display intermediate sensitivitytoward alpha-toxin (108), probably owing to the presence ofbinding sites of lower affinity (75).The only detailed binding study involving radiolabeled

toxin was published in 1976 by Cassidy and Harshman (40).Their tracer was radioiodinated to high specific activity butunfortunately retained only 10% of the original hemolytictiter. The data indicated that rabbit, but not human, eryth-rocytes express a limited number of surface receptors foralpha-toxin. Binding was optimal at 24°C and appearedirreversible. The authors concluded that presence of thesereceptors was responsible for the high sensitivity of rabbiterythrocytes. They concluded that at high concentrations,the toxin could bind via another, unspecific interaction toresistant cells such as human erythrocytes. Since ring struc-tures could be detected only on cells lysed at high toxinconcentrations, the possibility was considered that mem-brane damage invoked at high and at low concentrationsmight occur via different mechanisms. This in turn wouldplace the significance of hexamer formation in question.

Subsequently, Phimister and Freer (126) performed bind-ing experiments involving iodinated and functionally intacttoxin. They reported a uniform binding of approximately20% over the range of concentrations tested. These resultsthus did not support the receptor concept. Binding studiesconducted in our laboratory but using the enzyme-linkedimmunosorbent assay (ELISA) also generated no evidencefor the existence of receptor on rabbit erythrocytes (128).However, the sensitivity of the latter method would not havepermitted a small number of specific binding sites to bedetected.A very recent study has resolved the receptor controversy

by showing that the major results and conclusions of Cassidyand Harshman (40) were in fact correct. Hildebrand et al.labeled alpha-toxin to high specific activity with retention offunctional activity. Their binding studies demonstrated thatrabbit erythrocytes express an average of 1,500 to 2,000high-affinity, specific binding sites for alpha-toxin (76). Toxinbinds to target cells via two distinct types of interaction. Atlow concentrations, the toxin binds exclusively to the spe-cific sites; half-maximal binding occurs at a concentration of2 nM. At high concentrations (.200 nM), the toxin binds tomembranes via a nonspecific, absorptive interaction. Thisaccounts for the lysis of human erythrocytes and damage toliposomes. Autoradiographic analysis of the toxin bound toeither human or rabbit cells showed that hemolysis is inev-itably associated with the formation of hexamers; hence,

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from


MICROBIOL. REV.738 BHAKDI AND TRANUM-JENSEN

oligomerization is a prerequisite for pore formation at lowtoxin doses. Straightforward calculations revealed that he-molysis of rabbit cells was inducible through binding of veryfew hexamers, so that each hexamer might be forming apore. The very small number of oligomers generated at lowtoxin concentrations would render their detection by elec-tron microscopy virtually impossible, explaining why Cas-sidy and Harshman (40) could not detect rings on rabbit cellslysed with low toxin doses. Alpha-toxin binds to cells andforms hexamers at low temperature (0°C); however, thekinetics of both processes are slower (75).The nature of the specific binding site remains unknown.

Its small numbers (40, 76) render a proteinaceous naturemost probable. Our attempts to identify the binder moleculeby using ligand blotting, affinity chromatography, or cross-linking have uniformly failed. In view of the small number ofbinding sites, it seems improbable that they would be repre-sented by gangliosides (as proposed by Kato and Nakai [90])or by the anion antiporter (as proposed by Maharaj andFackrell [104]). As discussed above, the data of Maharaj andFackrell (104) could indicate a glycoproteinaceous nature ofthe binder molecule. Another potentially interesting obser-vation made by Kato et al. was that flavin mononucleotideapparently binds to and blocks the toxin-binding site onrabbit erythrocytes (91). This finding, if confirmed, couldopen new possibilities for isolation of the acceptor with theuse of affinity columns. Specific binding sites have recentlyalso been detected on human platelets and monocytes (75).The binding data explain why consumption of alpha-toxin

is generally low. When it is applied at high concentrations,binding occurs via the intrinsically inefficient adsorptiveinteraction. At low concentrations, binding is limited to thesmall number of specific sites expressed by certain cellsonly. Hence, it is comprehensible that alpha-toxin willgenerally be able to spread over relatively long distances todamage susceptible cells remote from its site of production.

MODE OF PORE FORMATION

We have proposed that alpha-toxin binds in monomerform to membranes and that hexamerization occurs upontheir subsequent collision during lateral diffusion in thebilayer (129). In support of this contention, application ofalpha-toxin at low doses to rabbit erythrocytes at 0°Cresulted in binding without substantial hexamer formationand without lysis. After washing and incubation of the cellsat 37°C, hexamerization accompanied by lysis occurred(129). Because unbound toxin had been washed away beforecells were warmed, hexamer formation must have been dueto oligomerization of the membrane-bound monomers.Blomquist and Thelestam (31) also treated adrenocortical Y1tumor cells with toxin at 0°C and noted that the membrane-bound toxin molecules exhibited different properties fromthe lytic hexamers forming at 37°C. Thus, for a short timeafter binding at low temperature, cell-bound (probably mo-nomeric) toxin could be inactivated by incubation withpolyclonal antibodies or by trypsin treatment. These prop-erties were lost after incubation of the cells at 37°C.The molecular events underlying pore formation have not

been elucidated, and a purely hypothetical model is pre-sented in Fig. 2. Each toxin molecule is proposed to containtwo domains (at and ,B) that display affinity for each other.Binding to a lipid bilayer serves to orient the molecules suchthat these domains are in the optimal position to interlockwhen two molecules collide during lateral movement in themembrane plane. Dimer formation could be accompanied by

Ub I Ub\W b W ___

FIG. 2. Proposed model for assembly of alpha-toxin hexamers.Each toxin molecule is considered to contain two domains (a and,B)that display high affinity for each other. Occupancy of each bindingsite (a and,) may trigger flipping-in of a respective and distinctamphipathic domain (a and b). The oligomerization process endswhen ring-structured hexamers have formed in which all reactivesites (a and P) are mutually occupied. The bulk of the toxinhexamers remain located outside the bilayer.

insertion of amphiphilic P-sheet domains into the bilayer. Insolution, toxin molecules are nonoriented, so the probabilitythat two molecules will collide in the orientation required totrigger unfolding is minimal. In fact, hexamers do formspontaneously, but slowly, in aqueous solution. The primaryfunction of binding sites on membrane surfaces will be toorient the toxin molecules uniformly so that they are primedfor unfolding when collision occurs. The oligomerizationprocess ends when ring-structured hexamers have formed inwhich all reaction sites (a and I) are mutually occupied.What is the fate of integral membrane components that are

displaced from their original locations? Theoretically, twopossibilities exist. First, lipids and integral membrane pro-teins might be expelled from the membrane and released asmicelles or small aggregates into the aqueous phase. Alter-natively, these components may be forced laterally apart.We have treated isolated erythrocyte membranes with lowand high doses of alpha-toxin and searched for lipids in themembrane supernatants. In no case could any expulsion oflipid ever be detected (15). Therefore, we believe thatintegral components are forced aside and that liberation ofintegral membrane constituents into the environment shouldgenerally not occur. It may be noted, however, that hexa-mers generated at high toxin concentrations tend to aggre-gate and small membranes blebs can form upon prolongedstanding that are released as microvesicles crowded withhexamers. This is a late phenomenon and is not related to theprimary mode of membrane damage.

STRUCTURE AND COMPOSITION OF PORES

First, we consider whether pores are uniformly composedof hexamers and whether membrane constituents may con-tribute towards their structure.


http://mm

br.asm.org/

Dow

nloaded from



A B

_Lp.

...... .... .

4

a b C aFIG. 3. SDS-PAGE analysis of rabbit erythrocyte membranes

treated with alpha-toxin (lane a), native alpha-toxin (lane b), andcontrol, untreated membranes (lane c). (A) Protein stain (Coomassieblue); (B) Western blot with a monoclonal antibody against alpha-toxin. Arrowheads indicate positions of toxin hexamers and mono-

mers. Only toxin hexamers and monomers are detectable on themembranes.

When toxin-treated membranes are solubilized with milddetergents, the toxin elutes upon molecular sieve chroma-tography as a symmetrical peak corresponding to the hex-amer form (62). Isolated hexamers can be reincorporatedinto lecithin liposomes. They present the typical cylindricalappearance in the electron microscope and generate a singleprotein band corresponding to the alpha-toxin monomer inSDS-PAGE (62). These data indicated that hexamers repre-

sent the predominant form of toxin pores and that membraneconstituents do not participate in forming the channel. Themajor assumption upon which these conclusions rest are (i)that lower-molecular-weight oligomers, if they exist, shouldalso remain intact in mild detergents, and (ii) that detergentwill not dissociate any bound membrane proteins. Althoughvalidation of these assumptions has not been possible, thefact that toxin alone could damage protein-free liposomesindicated that the straightforward hexamer concept could becorrect.Another method of analyzing the physical states of alpha-

toxin is to solubilize membranes with SDS at ambienttemperature and run gels in the cold. Toxin hexamers are notdissociated appreciably unless heated in SDS (41, 62), andthe electrophoresed toxin can be detected by autoradiogra-phy or by Western immunoblotting. All studies involvingthis approach have detected monomers and hexamers only(Fig. 3). Hence, if intermediate forms exist, they are notstable in SDS. Evidence for additional association of mem-brane proteins has never been obtained, but again, theseinteractions might be disrupted by the detergent.One unresolved puzzle relates to the fate of the specific

binding sites during hexamerization. There is no evidencethat the number of toxin hexamers increases with time at lowtoxin doses; i.e., the number of hexamers formed is fixedrelative to the number of binding sites (76). This excludes acatalytic action of the acceptor, and, by inference, theacceptor is likely to remain associated with the hexamerizedtoxin. Because the number of binding sites per cell is small,isolation of pores generated at low toxin concentrations(e.g., 100 to 500 ng/ml) poses a technical challenge. Such an

undertaking may nevertheless prove worthwhile, since thesepores could differ in structure and functional properties from

those generated following nonspecific absorption to themembrane at high toxin concentrations.Second, we consider the question regarding the structure

and orientation of hexamers in the membrane. Related tothis is the question of whether generation of a hexamer willinevitably result in the production of a pore.Thelestam et al. (149) and Harshman et al. (70) have

reported that membrane-bound alpha-toxin can be labeledwith an apolar photolabel, thus corroborating the contentionthat polypeptide domains become inserted into the mem-brane. Photolabeling may become a useful tool for identify-ing the membrane-inserted domains of the alpha-toxin mol-ecule in the future. In the electron microscope, toxinhexamers are seen as relatively thick-walled cylinders ap-proximately 4 nm high and 10 nm wide, apparently harboringa central pore of 1 to 2 nm in diameter (62). The cylinders areperpendicular to the membrane plane. Functional studieshave indeed shown that toxin-treated membranes harborpores of 1 to 2 nm in effective diameter. Could such defectsin the lipid bilayer be due to egress of molecules via leakypatches around the cylinder, i.e., between inserted proteinand lipid, as proposed by Bashford et al. (9)? It should berealized that the interaction between toxin and lipid must bevery tight, since it can be disrupted only by detergents (25,26, 62). We see no reason to assume that large, hydrophilicmolecules should be able to overcome such a barrier, and wewill therefore return to the concept of a pore passing throughthe center of the hexamer. A transmembrane leak would begenerated if each hexamer inserted sufficiently into thebilayer and if the channel extended through the entire lengthof the hexamer in an open configuration. Regarding thedepth of insertion, electron microscopy indicates that thebulk of the toxin hexamer is located outside the lipid bilayer.Toxin cylinders appear to project 3 to 4 nm from themembrane surface into the extracellular (aqueous) phase(62). Since the thickness of the hexamer wall is 3 to 3.5 nm,an approximate molecular weight of 150,000 to 200,000 canbe calculated for the extramembranous domain. Althoughthis molecular weight estimate is crude, it does indicate thatonly a minor part of the hexamer is left to form theintramembranous pore domain. The walls of this domainmay be quite thin, and the possibility that the hexamer doesnot penetrate through the entire bilayer must also be consid-ered. Should this be true, diverse membrane factors mightobviously influence or even inhibit the formation of pores.

Recently, Olofsson et al. (118, 119) reported low-resolu-tion structural analysis of toxin pores by using image proc-essing of electron micrographs. They examined single oligo-mers randomly distributed in platelet membranes (119), aswell as oligomers that had accumulated into two-dimensionalcrystals in artificial lipid bilayers (118). Unexpectedly, thetwo oligomer preparations exhibited quite different struc-tures. The single particles had smaller internal diameters of7 nm as opposed to 10 nm in the crystals. Remarkably,whereas a hexamer structure evolved from analysis of singleparticles, the crystals showed either four or eight proteinpeaks. Moreover, a lid that partially occluded the centralchannel appeared to be present in the crystalline oligomers.Occlusion of the central pore was not observed in the toxinhexamers formed on platelets. Further work is obviouslyneeded to clarify these apparent discrepancies. At this stage,it seems reasonable to conclude that oligomers forming onbiological membranes (e.g., platelets) do represent hexamerscontaining a water-filled channel in their center.Two findings currently indicate that, whereas hexamer

formation is required for membrane damage to occur, not all

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from



hexamers necessarily need form pores. First, high concen-

trations of Ca2" can apparently close toxin pores in cells (9,72) and planar lipid membranes (114). The mechanism un-

derlying this process is unknown. Although these observa-tions are interesting from a theoretical point of view, thebiological significance of the phenomenon is unclear. Bash-ford et al. postulate a protective role of extracellular Ca21 on

nucleated cells in a biological setting (9), but we do not sharetheir opinion. As will be discussed below, susceptible cellssuch as platelets, endothelial cells, and monocytes are

rapidly permeabilized by low concentrations of alpha-toxinin the presence of physiological Ca2 levels and, indeed,calcium influx through the pores probably represents themost important trigger for many detrimental cellular reac-

tions.A second finding that cautions against the generalization

that hexamer formation is always accompanied by mem-

brane permeabilization is that alpha-toxin binds to certaincells with formation of hexamers, but without invokingmembrane leakiness. This has been observed with humangranulocytes treated with relatively high toxin doses (20,ug/ml) at 22 to 37°C (75). Obviously, the hexamers do notproduce pores in these cases. The reasons for the intrinsicresistance of these cells remain unknown. Although thepossibility that granulocytes can rapidly remove the hexa-mers from their surface springs to mind, pilot experiments totest this hypothesis have uniformly yielded negative results(75).

In summary, the available data indicate that hexamerformation is absolutely required for membrane permeabili-zation to occur and that pores probably traverse the interiorof the hexameric structures. On the other hand, unclarifiedmechanisms that can sometimes counteract the permeabiliz-ing action of the hexamers appear to exist.

FUNCTIONAL PROPERTIES OF TOXIN PORES:DAMAGE TO LIPOSOMES AND

HEMOLYTIC CAPACITY

In membranes of susceptible cells, toxin hexamers gener-

ate nonselective pores, of 1 to 2 nm in effective diameter,that permit free passage of ions and low-molecular-weightmolecules such as nucleotides. The molecular weight cutofflies in the region of 1,000 to 4,000. Dextran 4 (Mr 4,000) doesnot pass through the pores and can be used to prevent

osmotic swelling of cells. Hemolysis is totally inhibited bythis dextran when it is used at a concentration that counter-

balances the colloid osmotic pressure exerted by hemoglobin(21). Toxin pores in planar lipid bilayers have similarly beencharacterized by Menestrina and coworkers as essentiallynonselective, water-filled channels with diameters of 1 to 1.5nm (10, 114). The pores are voltage insensitive and remainopen irrespective of the membrane potential. Once formed inplanar membranes, toxin pores are not destroyed by trypsintreatment. However, as mentioned above, high Ca2+-con-centrations close the pores in a reversible fashion (114).

Several recent investigations in which liposomes were

used as targets indicate that cholesterol (58), unsaturated

fatty acids (80, 81), and phosphatidylcholine (164) augment

the efficiency of toxin attack. Since incubation with phos-phatidylcholine led to toxin inactivation, it has been pro-

posed that choline-containing phospholipids may be in-

volved in binding of the toxin (164). Toxin hexamerization

was also found to be induced by deoxycholate (17), wherebymarked and unexplained differences in the extent of oligo-

mer formation have been noted to be dependent on the toxin

and deoxycholate charge (15). Raff et al. (127) observedinactivation of alpha-toxin by hydrocortisone and methyl-prednisolone, which might also have been due to hexamerformation induced by the sterols.

In a recent study, alpha-toxin was applied to asymmetricalplanar lipid bilayers lined on one side with bacterial lipopoly-saccharide and on the other with phospholipids (135). Ap-plication to the phospholipid face but not to the lipopolysac-charide face resulted in pore formation. Thus, severalexplanations have now emerged for the early finding thatalpha-toxin does not damage bacterial cell membranes (12):lack of cholesterol, paucity of exposed phosphatidylcholine,and shielding by lipopolysaccharide molecules.

Determination of hemolytic titer represents an importantmeans of checking the quality of a toxin preparation. Thestandard procedure is to add a suspension of rabbit erythro-cytes to serially diluted toxin. The reciprocal of the dilutioneliciting 60% lysis within 1 h at room temperature gives thenumber of hemolytic units (HU), which can be expressed permilligram of protein. The specific activity of purified alpha-toxin is in the range of 40,000 HU/mg of protein, whenassessed by addition of 1 volume of a 2.5% erythrocytesuspension (2.5 x 10' cells per ml). The hemolytic titer ishigher when incubation times are prolonged and reaches50,000 to 100,000 HU/mg after 4 h at room temperature. It isnoteworthy that toxin binding and hemolysis are most effec-tive at room temperature. Many of the early studies wereconducted with toxin preparations that exhibited markedlylower activity, and the possibility that either a substantialpart of the toxin was inactive (e.g., in aggregated form) orthe preparations were contaminated must be considered.

STRUCTURE-FUNCTION RELATIONSHIPS

The subject of structure-function relationships is currentlya problematic area in alpha-toxin research. Several studiespurport to have identified molecular domains required formembrane binding, as well as domains that are involved inoligomerization and pore formation. Other studies haveattempted to identify amino acid residues that are essentialfor toxic function by using biochemical modification meth-ods.Harshman et al. generated monoclonal antibodies against

N- and C-terminal peptide domains (68, 73). Because anantibody directed against an epitope located in the C-termi-nal fragment inhibited the binding of alpha-toxin to cells, thebinding region was proposed to reside in the C-terminal halfof the molecule. An antibody against the N-terminal half didnot inhibit binding, but inhibited oligomerization of thetoxin. On the basis of radioiodination experiments (39),Cassidy and Harshman suggested that a tyrosine residuewithin the N-terminal domain is essential for the oligomer-ization process. Unfortunately, antibody molecules, with amolecular mass greatly exceeding that of alpha-toxin, maynot be sufficiently accurate probes to clarify structure-function relationships, and antibody binding could pose asteric hindrance to an interaction involving a molecularregion distinct from the epitope. Ultimately, a binding do-main must therefore be identified by using more rigorouscriteria, e.g., competition of binding of the native molecule.However, successful competition studies have not beenreported. Alternatively, binding of the respective polypep-tide domain or toxin fragment might be demonstrated di-rectly. Binding data are available in one case from experi-ments by Blomquist and Thelestam (32, 33). These authorsisolated a naturally occurring fragment of alpha-toxin, iden-

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from



tified as the C-terminal half, that bound to cells (33) andapparently possessed full hemolytic activity (32). These datathus indicated that only approximately half of the alpha-toxin molecule is required for pore-forming activity in eryth-rocytes, and the N-terminal half would not be necessary foroligomerization. Although limited proteolysis with trypsinalso generates fragments of similar size by cleaving bonds inthe center of the molecule, the ensuing C-terminal fragmentof molecular weight 17,000 (the 17K fragment) is entirelydevoid of binding or lytic properties. A remarkable andunvoiced conclusion emerging from these considerations isthat a small stretch of amino acid residues present in the 18Kfragment of Blomquist and Thelestam (32, 33) but absent inthe C-terminal 17K trypsin fragment must be involved inboth the binding and oligomerization processes.

In summary, there is currently some reason to assume thatthe binding domain of alpha-toxin is located in the C-termi-nal half of the molecule. It is not known whether the domainis represented by a linear sequence or whether several partsof the molecule participate in forming a conformationalepitope.

Blomquist and Sj0rgen (30) have reported that differentmonoclonal antibodies vary in their respective abilities toinhibit the hemolytic, dermonecrotic, and lethal effects ofalpha-toxin. For example, one antibody did not neutralizethe hemolytic or dermonecrotic effects, but did abrogatelethality. Hence, the authors believe that different biologicalactivities of alpha-toxin are carried or evoked by differentmolecular regions. This contention is provocative because itimplies either that the toxin has different binding sites fordifferent cell targets or that the various toxic effects are notuniformly due to hexamer transmembrane pore formation.In a similar context, Blomquist et al. reported that theC-terminal, naturally occurring 18K fragment of alpha-toxin,although hemolytic, was not cytotoxic to Y1 cells or lethal inmice (29, 32, 33). This fragment appeared to bind to andblock the receptor on Y1 cells without incurring any damage,and pretreatment of the cells with the fragment renderedthem refractory to subsequent action of native toxin (33).Further, the fragment appeared to oligomerize on Y1 cells toform noncytotoxic dodecamers (33). These very interestingobservations have not been corroborated. It is somewhatdisappointing that, in the face of the importance of theirfinding, the authors did not conduct a binding competitionstudy with their toxin fragment. It should be mentioned thatWatanabe and Kato earlier claimed that a 17K, C-terminaltoxin fragment, although devoid of hemolytic activity, waslethal in mice (163). However, the N-terminal sequenceobtained for their fragment (93) could not subsequently befitted into any part of the complete amino acid sequence thatbecame available thereafter. This example underlines theneed for rigorous criteria to be applied in any study that isaimed at elucidating structure-function relationships at amolecular level.

Fackrell and Hebert (54) reported that treatment of alpha-toxin with an arginine-modifying reagent destroyed its he-molytic activity. The concentration of the modifying reagentused was very high, and it was unclear whether the modifiedtoxin failed to bind or whether the oligomerization step wasinhibited. More informative studies involving chemical mod-ifications were recently performed by Menestrina and co-workers (43, 124). First, they found that modification of twoor more histidine residues with diethylpyrocarbonate led toboth defective binding and defective oligomerization. Theirtitrations indicated that three histidine residues are avail-able for reaction with diethylpyrocarbonate in the toxin

monomer, whereas only one residue is reactive in the mem-brane-embedded hexamer. Detergent-extracted hexamersappeared to present two reactive histidine residues. Theoriginal sequence data of Gray and Kehoe (66) reportedthree histidine residues at positions 35, 144, and 259; anotherresidue is present at position 48 (94a). At present, it remainsunknown which of the histidine residues are involved in thebinding and oligomerization processes, respectively.

In another study, Cescatti et al. (43) found evidenceindicating that lysine residues line the channel and influencethe electrical properties of the pore. Studies involving site-directed mutagenesis should be useful for more preciseidentification of the relevant amino acid residues that takepart in the binding and oligomerization processes in thefuture.

EFFECTS OF PORE FORMATION ON TARGET CELLS

Cell Death

If a single pore of 1 to 2 nm in diameter is formed and notremoved from the plasma cell membrane, cell death can beexpected to ensue as a result of rapid egress of vitalmolecules such as ATP and the loss of the characteristic"milieu interieur" that is required to sustain metabolicprocesses. Permeabilized erythrocytes undergo irreversibleosmotic swelling, and hemoglobin is finally liberated as themembranes rupture. Designation of the toxin as a hemolysinis therefore justified. The term "cytolysin" is more proble-matic since nucleated cells and platelets do not really lyse,but simply swell because their content of cytoplasmic mac-romolecules is lower (19, 23). Pore formers such as alpha-toxin therefore do not primarily cause liberation of macro-molecular cytoplasmic components. Pore formation anddeath of nucleated cells can conveniently be registered byconventional dye exclusion tests, by measuring the uptake ofa fluorescent dye such as propidium iodide (19), or bymeasuring ATP leakage (20, 22, 23).Death of a particular cell type can, of course, have some

primary biological consequences as a direct result of loss ofthe respective cell function. For example, monocytes andmacrophages are effectively killed by alpha-toxin, and thisconceivably cripples the phagocytic defence system in tis-sues (22, 111). Perfusion of rabbit lungs with very low dosesof alpha-toxin causes death and detachment of endothelialcells from the basal membrane, i.e., breakdown of theendothelial permeability barrier with consequent leakage ofmacromolecules and water from the vascular space into thealveoles (134). In a related context, Vann and Proctor (160)have found that uptake of S. aureus leads to the death ofendothelial cells, possibly owing to intracellular release ofalpha-toxin.

Secondary EffectsUntil quite recently, a gap yawned between basic studies

dealing with the primary mechanism of alpha-toxin actionand studies describing cellular reactions and pathophysiolog-ical effects. Were such complex phenomena as plateletactivation, or lethal effects in general, causally linked tomembrane pore formation and, if so, via which mechanisms?Today, there is increased awareness that the generation oftransmembrane pores will trigger pathophysiologically im-portant, secondary cellular reactions (25, 26). Two factsprovide the basis for this. First, the pores are too small topermit egress of cytoplasmic proteins; therefore, molecular

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from


MICROBIOL. REV.742 BHAKDI AND TRANUM-JENSEN

14001Pg/ml

1200

1000-

800.

600.

400 .

x 6KetoPGFia

0 TXB2

2000y

PAP KHH -Ca"-fr.es,EGTA(SMM)(mmHg) I80180j a-Toxin (5pi)64-

56-

4840322416_

0.

KHHB |KHMB

° to (min) O 5 10 S min)

FIG. 4. Perfusion of alpha-toxin through isolated and ventilated rabbit lungs induces Ca2"-dependent generation of prostanoids and anincrease in pulmonary artery pressure (PAP). The figure shows the course of PAP and concentrations of thromboxane B2 (TXB2) and 6-ketoprostaglandin Fl. (6KetoPGFi,) in the recirculating perfuion fluid after application of 5 ,ug of alpha-toxin in two different lungs in the presenceand absence of intravascular calcium. The dashed lines in the right-hand panel indicate the change of perfusion fluid from calcium-free[ethylene glycol-bis(j3-aminoethyl ether)-N,N,N',N'-tetraacetic acid (EGTA)-containing] Krebs-Henseleit hydroxyethylamylopectine buffer(KHHB) to normal calcium-containing KHHB. Note the calcium-dependent rise in PAP accompanied by the appearance of eicosanoids.Reproduced from the Journal of Clinical Investigation (133) by copyright permission of the Rockefeller University Press.

machineries for diverse cellular processes remain intact.Leakage of smaller molecules such as ATP, although fairlyrapid, still requires several minutes. During this time, evencells that are dying can mount a series of pathophysiologi-cally relevant reactions. Second, calcium ions diffuse rapidlyalong the steep extra- to intracellular concentration gradient,and they probably represent the most important singletrigger of secondary reactions. Of these secondary reactions,the following have been studied in some detail.

(i) Stimulation of eicosanoid production. Calcium ionstrigger arachidonic acid metabolism via their participation inthe activation of membrane-bound phospholipase. Today,the fact that cells attacked not just by alpha-toxin, but by allpore formers studied, respond with rapid production ofprostaglandins or leukotrienes, or both, is therefore notsurprising (25, 26). When these findings were first reported,however, they provided an unsuspected link between thephenomenon of membrane damage and a complex cell-biological reaction (133). The key role of calcium influx wasfirst documented in experiments utilizing alpha-toxin (133,141, 143, 144) and was subsequently extended to othercytolysins.The pattern of eicosanoid production varies qualitatively

and quantitatively among different cell species. Each lipidmediator in turn elicits a typical spectrum of cellular reac-tions, so that long-range consequences may ensue in acomplex system (e.g., whole organs). For example, culturedbovine endothelial cells release large amounts of prostacy-

clin and thromboxane following attack by very low doses ofalpha-toxin (143). Marked production of thromboxane isadditionally noted in isolated, perfused rabbit lungs. Throm-boxane is known to exert a vasoconstrictive effect on lungvessels. When alpha-toxin is perfused through isolated andventilated blood-free rabbit lungs, it evokes a steep rise inpulmonary vascular resistance (133). This is accompanied bya protracted breakdown of the vascular permeability barrierowing to additional, direct cytotoxic effects of the toxin onthe endothelial cells (133, 134). As a result, pulmonaryedema develops (Fig. 4). These experiments have shownthat a bacterial cytolysin can thus provoke pulmonary dys-function and a pathological status resembling adult respira-tory distress syndrome.

Alpha-toxin can also trigger production of leukotrienes inpolymorphonuclear granulocytes (144). Owing to the intrin-sic resistance of these cells, however, the doses required toevoke this effect are very high. Eicosanoid production inother more susceptible cells (e.g., monocytes and macro-phages) has not yet been studied.

(ii) Secretory processes. Calcium ions represent a majortrigger for secretion, and alpha-toxin was originally intro-duced as a tool to study the minimal requirements forexocytosis (3). Neuroblastoma PC 12 cells permeabilizedwith the toxin mounted a secretory response upon additionof micromolar amounts of calcium ions to the medium. Arequirement for ATP, Mg, or GTP could not be discerned. In

140lpg/ml

x 6KetoPGF,a

0 TXB2

1200

1000,

800-

400

200

0

PAP(mmHg)

4&

40-

32

24

18

O0

KHHB

a-Toxin (Sag)


http://mm

br.asm.org/

Dow

nloaded from



U)V-6

4)

Q0

xCn

.C)

E._1

c._

E20

1.5

1.OjF

0.51-

@ 1--W.1 1.50)0

0.,- 1.0

-10)

.' 0.5E

3

)--

o0

0.0

.... . . ..- --.X * -. -l * * - Ia10 100 1000 10000

a-toxin (ng/mI)

0 50 100 150 200

IgG concentration (,Ig/mi)FIG. 5. Alpha-toxin induces assembly of prothrombinase com-

plexes on human platelets. (A) Platelets were treated with variousamounts of alpha-toxin in the presence of 3 mM CaCl2 and pro-

thrombin for 10 min at 22°C. The prothrombinase reaction was

started by addition of factor Xa (33 pM). After 10 min, aliquots were

removed and assayed for thrombin. Total prothrombinase activity(a) was determined by lysing platelets through three cycles offreezing and thawing. (B) Platelets were incubated with 0.1 pg ofalpha-toxin per ml in the presence of irrelevant immunoglobulin G(IgG) or specific anti-factor V immunoglobulin G, and the ensuingprothrombinase activity was measured. The negative control (V)received no toxin. This experiment identified platelet-derived factorV as the central element responsible for the appearance of pro-thrombinase activity. Reproduced with permission from the Journalof Biological Chemistry (7).

this case, therefore, calcium ions appeared to represent thesole requirement for exocytosis (3).

Calcium-dependent secretory responses were subse-quently documented to occur in platelets, which rapidlysecrete large amounts of platelet factor 4 and factor Vfollowing attack by alpha-toxin (7, 23). As originally notedby Siegel and Cohen (136), the platelets do not lyse, butundergo a swelling reaction and leak ATP. Secreted factor Vassociates with the platelet surface, forming nuclei for as-sembly of prothrombinase complexes (Fig. 5). This consti-tutes the major pathway responsible for the procoagulatoryeffects of alpha-toxin (7). In this connection, earlier obser-vations on "lysosomal labilization" in cells treated withalpha-toxin (83, 84) may have been due to secretion oflysosomal constituents. Exocytosis will probably occur in alarge variety of cells permeabilized by alpha-toxin.

(iii) Cellular contractile dysfunction. Calcium ions causecontraction of actin-myosin filaments, and relaxation is an

active process requiring ATP. Since toxin-permeabilizedmembranes leak ATP and cells are flooded with calcium,subsequent dysfunction of the cellular contractile apparatusmight be anticipated. Indeed, when Suttorp et al. (142)treated confluent endothelial-cell monolayers with low dosesof alpha-toxin, they noted that the endothelial cells con-tracted, leading to formation of intercellular gaps that per-mitted free passage of macromolecules across the originallyintact monolayer (Fig. 6). Similar processes, if occurring invivo, would promote the development of edema.

(iv) Activation of endonucleases. Programmed cell death iswidely believed to be caused by activation of endogenousendonucleases, and calcium ions have been implicated as amajor trigger for this process. Endonuclease activation hasbeen proposed to represent one of the killing mechanisms ofcytotoxic lymphocytes. One study by Hameed et al. (67)reported that DNA degradation also occurred in cells per-meabilized by alpha-toxin, and endonuclease activation wasproposed to represent a general mechanism which is opera-tive in cells damaged by pore-forming proteins. The obser-vation, although interesting, requires confirmation becausethe toxin preparation used by the authors was very impureand may have contained endonucleases. Several hypothesescurrently accommodate regulatory factors (such as proteinkinase C) that could suppress endonuclease activation in thepermeabilized cells (109). At present, the contention thatendonuclease activation occurs in toxin-treated cells as anearly event appears interesting but requires more extensivestudy.

(v) Release of cytokines. Human monocytes are highlysusceptible to damage by alpha-toxin (22). Current evidenceindicates that alpha-toxin induces rapid processing and se-cretion of interleukin-li if the interleukin precursor hasaccumulated intracellularly. Toxin-triggered release of inter-leukin-l, was very rapid (complete within 60 min) andoccurred in the presence of cycloheximide and actinomycinD (Fig. 7). Alpha-toxin might thus synergize with variousstimuli that trigger interleukin-1 synthesis to accelerate andaugment release of the cytokine (22).

(vi) Membrane alterations. Can insertion of an alien pro-tein into the bilayer lead to reorientation of lipid compo-nents? According to one study, this may be the case.Schneider et al. (131) detected enhancement of the flip rateof lysophosphatidylcholine from the outer to inner mem-brane leaflet of erythrocytes in correlation with alpha-toxin-induced lysis. By extrapolation, lipid components couldpossibly flip from the inner to the outer monolayer as well. Inthis context, indirect evidence exists for a reorientation ofmembrane constituents in toxin-treated platelets, whichwere observed to bind factor Va in an augmented fashion.Binding of factor Va to the platelet surface involves nega-tively charged phospholipids, and augmented binding mayhave been due to flip-flop of phosphatidylserine from theinner to the outer leaflet. Analogous findings have been madefor platelets permeabilized with complement (166, 167).

(vii) Other effects. Gemmell et al. (63, 64) reported thataddition of alpha-toxin to human serum resulted in comple-ment consumption and, consequently, in reduction in op-sonic activity of the serum. The relevance of this finding isunclear since the concentration of alpha-toxin used to pro-duce immune complexes was probably very high (of note,their toxin preparation exhibited a specific activity of only1,200 HU/mg of protein). In another study, we did not detectsignificant complement consumption as a result of additionof highly purified, monomeric toxin to serum (16). We havealso not confirmed the report of Gemmell et al. (63) that

A

Bcontrol IgG

I-

1-F

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from



ilp oerv a ;,? ; r4 -4,Sv

.,

*.X~2

.4~~~~~

**f; 'rs S

~~~~7~~~~~~~A ~ l

C Fwri-S C,

FIG. 6. (A and B) Sealed endothelial monolayer of pig pulmonary artery grown on gelatin-coated coverslips and exposed for 90 min to ahydrostatic pressure of 10 cm of H20. Examination by phase-contrast microscopy (A) and scanning electron microscopy (B) shows thatvirtually all interendothelial junctions (intercellular spaces) are closed. (C and D) Endothelial monolayer exposed for 90 min to 1 ,ug ofalpha-toxin per ml. Note large interendothelial gaps (arrows) that are seen by both phase-contrast microscopy (C) and scanning electronmicroscopy (D). Bars, 20 ,m. From Suttorp et al., Am. J. Physiol. 255:C368-C376 (1988). Reproduced from the American Journal ofPhysiology (144) with permission from the authors and the publisher.

alpha-toxin directly enhances the phagocytic killing capacityof neutrophils (15).

Petrini and Moilby (125) reported activation of T lympho-cytes ly alpha-toxin. Further investigations in this area, inparticular with regard to dose-response relationships, arewarranted.Harshnian and colleagues reported disruption of myelin

sheats in isolated rabbit nerves (139) and in mouse brain invivo and, in vitro (71) by alpha-toxin. Chan and Lazarovici(44) have observed a calcium-independent phosphorylationof several proteins including myelin basic protein aftertreatment of myelin with alpha-toxin. The same authorsfound that incubation of PC 12 cells with subcytotoxic dosesof alpha-toxin resulted in reduction of epidermal growthfactor receptor affinity (99). Future work should aim atclarifying the underlying mechanisms of these processes andtheir pathogenic relevance.According to Kato et al. (89, 94), alpha-toxin also exerts

an inhibitory effect on isolated AMP-dependent proteinkinase. The toxin seems to compete with cyclic AMP for abinding site on the regulatory subunit of the enzyme. Thebiological significance of this finding is unclear since alpha-toxin is normally not able to reach a cytoplasmic target.

However, in. conjunction with the observation that flavinmononucleotide inhibits hemolysis of rabbit erythrocytes(91), one might speculate that alpha-toxin interacts specifi-cally with nucleotide-binding sites on several proteins. Oneof these could possibly be identical with the biologicallyimportant acceptor molecule on cell surfaces.

(viii) Repair processes. Thelestam and Moliby reportedthat cultured cells could recover from toxin attack (152).This investigation raised some very interesting questionssince membrane permeabilization was apparently followedby regaining of bilayer integrity. The mechanism of mem-brane repair was not studied, and the extent to which othercells can deal with a limited number of toxin lesions isunknown.

Alpha-Toxin as a Biological Tool

A number of properties render alpha-toxin an excellenttool for controlled permeabilization of cell membranes. Thetoxin is very stable, and neither transport nor storage posesany problem. It is soluble in water or any aqueous buffer andis stable over a wide pH range. No ionic requirements havebeen noted for membrane binding or pore formation, and

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from



602o-

40

1LL

o'10-

0 20 50 100 200 500 1000 2000 5000ngImI It-toxin

FIG. 7. Action of alpha-toxin on cultured monocytes in the

presence of 10%o autologous serum. (A) Cellular ATP content,expressed as a percentage of untreated control levels. (B and C)

concentrations of interleukin-1p (IL-103 (panel B) and tumor necro-sis factor (TNF) (panel C) in the cell supernatants. The incubationperiod with alpha-toxin was 6 h at 37°C. Each plot depicts the results

obtained from one individual. Reproduced from Infection and

Immunity (22) with permission.

permeabilization can thus be performed in the presence of

chelating agents. Permeabilization is most efficient at room

temperature, but can also be performed at 4 or 37°C. The

toxin attaches exclusively to the plasma membrane and does

not reach intracellular organelles or cytoplasmic components.Pores created by the toxin do not permit egress of macromol-

ecules from the cytoplasm; hence, enzyme cascades and

cellular machineries (e.g., for secretion) remain intact over

extended periods. Controlled manipulation of the intracellularionic composition, as well as introduction of nucleotides, is

thus feasible. Alpha-toxin is increasingly being used as a tool

to study exocytosis, and a number of publications have

appeared in this area (3, 4, 8, 65, 77, 107, 110, 116, 132, 158,

161). The advantages afforded by alpha-toxin over other

permeabilizing agents have been emphasized (65).

BIOLOGICAL RELEVANCE OF ALPHA-TOXIN

To arrive at an appreciation of the possible biological

relevance of any cytolysin, it is necessary to identify the

cells that are preferentially attacked under physiologicalconditions (e.g., at low toxin concentrations and in the

presence of plasma proteins) and to study the pathophys-iological consequences thereof. It is now clear that alpha-

toxin will permeabilize any mammalian cell membrane if

applied at a sufficient dosage; therefore, the observation that

cells can be damaged in vitro is unsurprising and does not

itself shed light on the biological relevance of toxin produc-tion. For example, despite its classical designation as ahemolysin, the pathogenic potential of the toxin does not liein its capacity to lyse erythrocytes. The list of cells that havebeen used as targets is long, and readers are referred toreviews by Wiseman (168), Jeljaszewicz et al. (83), andThelestam and Blomquist (148) for compilations. Today, thefirst question to be raised when comtemplating the possiblein vivo relevance of any cytotoxic action of alpha-toxinrelates to the concentration required to evoke the respectiveeffects. If these are observed at toxin concentrations below1 ,ug/ml (ca. 50 HU/ml), the target cells most probablyexpress high-affinity binding sites and cytotoxic effectsshould also be expected to occur under physiological condi-tions. This assumption is underscored if the cytotoxic actioncan be shown to occur in the presence of plasma proteins.Among the well-documented examples of such highly sensi-tive cells are rabbit erythrocytes (49, 108), platelets andendothelial cells from several animal species (14, 23, 133,134, 136, 142, 143), murine macrophages (111), humanmonocytes (22), and adrenocortical tumor cells (31).

In an in vivo setting, the consequences of toxin liberationwill be governed on the one hand by the presence of suchhighly susceptible target cells and on the other hand byinactivating mechanisms in the host organism. A brief dis-cussion of the latter follows here.

Like other known pore-forming bacterial cytolysins, al-pha-toxin is an excellent immunogen. All healthy adults haveplasma antibodies against the toxin, attesting to the fact thatimmunogenic amounts are produced by staphylococci evenin the absence of clinically manifest infections. Indeed, titerdeterminations based on a neutralization test or ELISA havebeen used in some laboratories as a diagnostic aid in cases ofcertain chronic staphylococcal infections, such as osteomy-elitis and endocarditis, and in septicemias (53, 87). In addi-tion, alpha-toxin can be bound and partially activated bylow-density lipoprotein (18). A basic question hence arises:given the presence of these neutralizing mechanisms, howshould alpha-toxin be able to successfully attack even cellsthat are highly susceptible in vitro? This question is nottrivial. In classic diseases caused by bacterial toxins, such astetanus, diphtheria, and cholera, the presence of neutralizingantibodies is virtually synonymous with immunity. This isnot the case with alpha-toxin and other pore formers.

Theoretically, there are three basic ways by which inacti-vating mechanisms may be sidetracked or overrun. First,toxin may be released into an environment that lacks theinactivators (e.g., infected tissues). Second, a toxin might besecreted by bacteria that are cell adherent and thus beinaccessible to the inactivators. Third, toxin binding tosusceptible cells may occur more rapidly than the binding ofthe inactivators; hence, cells may be successfully attackeddespite the presence of the inactivators. The last possibilitydeserves some emphasis. Moderate concentrations of alpha-toxin (1 to 2,ug/ml) will be perfectly harmless to platelets ifthe toxin is preincubated for 1 min with normal human serumor with a neutralizing monoclonal antibody (Fig. 8). In theabsence of such a preincubation step, however, the toxinsuccessfully binds to and activates platelets suspended inplasma (23). An analogous phenomenon has been demon-strated for the attack of E. coli hemolysin on human granu-locytes (19). These examples serve to illustrate that antibod-ies and lipoproteins, although endowed with considerableneutralizing capacity when tested in a conventional mannerin vitro, may be unable to effectively protect highly suscep-

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from



A

t

..., \ .:.As . ...* ;a 4 t -.FIG. 8. Effect of neutralizing monoclonal antibody a4Cl (for

details, see reference 68) on platelet aggregation and ATP releaseinduced by alpha-toxin. (A) Alpha-toxin was preincubated with theantibody at a molar ratio of 1:1 for 2 min at 22°C and then added toplatelet-rich plasma (final toxin concentration, 2.5 p.g/ml); no effectsof the toxin were discerned. Addition of another 2.5,ug of toxin per

ml to the sample resulted in platelet aggregation and ATP release.(B) Toxin and monoclonal antibody were applied simultaneously inthe same dose as in panel A. In this case, the antibody could not

prevent platelet activation and aggregation. (C) Toxin was applied30 s before the monoclonal antibody; no protective effect of theantibody was observed. Reproduced from the Journal of Experi-mental Medicine (23) by copyright permission of the RockefellerUniversity Press.

tible cells against attack by bacterial cytolysins under phys-iological conditions.

Relevance in Animal Models

Two basic approaches may be undertaken to probe thesignificance of toxin production. The first is to demonstrateprotective effects of specific antibodies in a defined model ofanimal infection. The second is to demonstrate differences inthe virulence of bacterial strains exhibiting differences intheir ability to produce the toxin. A number of studies havegenerated positive data in both cases. Adlam and associatesinvestigated the development of staphylococcal lesions andthe role of alpha-toxin in lactating mammary glands inrabbits (2, 162). They demonstrated a deleterious effect of

purified alpha-toxin and showed that immunization withhighly purified toxoid (but not with beta-toxoid) reduced the

lethal edematous form of the disease, although abscess

formation was not prevented. Taubler et al. similarly ob-

tained evidence for a role of alpha-toxin in the pathogenesis

of subcutaneous lesions in mice (145). Van der Vijver et al.

(159) induced mutants of S. aureus and found that toxin-

deficient mutants were less virulent than parental strains.Kapral et al. (88) obtained evidence that alpha-toxin was

produced in intraperitoneal abscesses of experimentally in-

fected mice. These investigators detected hemolytic activitycorresponding to 100 to 1,000 HU/ml of alpha-toxin in thelesions. Hence, large amounts of alpha-toxin may indeed beproduced in staphylococcal abscesses.A major difficulty in earlier experiments stemmed from the

fact that production of several potentially important viru-lence factors in addition to alpha-toxin is under regulation bya common control element (1, 27, 47, 48, 115, 128, 172).Hence, ultimate proof for a pathogenic role of the toxincould be obtained only through the use of genetically engi-neered, isogenic bacterial strains differing solely in theexpression of the alpha-toxin gene. The few studies con-ducted under these premises have yielded positive results.O'Reilly et al. (120) inactivated the alpha-toxin gene clonedin E. coli by site-directed mutagenesis. After reintroductionof the active or inactive gene into nonlytic staphylococci,they showed that the toxin-producing strain was more viru-lent than the toxin-negative isogenic strain in a mouseperitonitis model. In a further study with isogenic strains,Patel et al. (122) similarly showed that alpha-toxin was amajor causative factor in the development of subcutaneouslesions in animal infections. Furthermore, Bramley et al.(34a) demonstrated that alpha-toxigenic bacteria survivedphagocytosis during the initial inflammatory response inintramammary infections. Jonsson et al. (86) mutagenized anS. aureus strain and examined the virulence of differentmutants by using a mouse mastitis model. They found thatmutants simultaneously defective in coagulase and alpha-toxin production expressed dramatically reduced virulence,whereas mutants lacking only one of these factors displayedintermediate virulence. Their approach, although not asprecise as the use of isogenic strains by Foster's group (120,122), thus also generated evidence for a pathogenic role ofalpha-toxin in the mastitis model. In sum, these experimentsuniformly indicate that alpha-toxin is truly relevant as avirulent factor in animal infections.

Relevance in Humans

It is presently impossible to prove that alpha-toxin repre-sents a pathogenetic factor in humans. Staphylococcal dis-ease is multifactorial and is usually due to the simultaneousproduction of several pathogenetic factors including proteinA, coagulase, and leukocidin. Attempts to compare theincidence of alpha-toxin production and relate this propertyto clinical symptoms would hardly be meaningful, and nosuch studies have been undertaken. Our bias for the patho-genic relevance of alpha-toxin is based on the knowledgethat (i) certain human cells including monocytes, endothelialcells, and platelets express high-affinity binding sites and areeffectively attacked by the toxin under physiological condi-tions and (ii) damage occurring to these cells triggers patho-logical sequelae. Hemostasis disturbances, thrombocytope-nia, and pulmonary lesions are often encountered in patientsduring severe staphylococcal infections.One way of indirectly demonstrating a significant role of

alpha-toxin as a virulence factor may be possible if applica-tion of specific hyperimmune globulins were found to betherapeutically effective. Human hyperimmune globulinsagainst alpha-toxin have been prepared, and they are able tosuppress all experimental toxic effects in vitro and in vivo(20) (Fig. 9). Patients with early diagnosis of S. aureussepticemia are candidates for therapy with such hyperim-mune globulins, but clinical trials with these antibodies haveyet to be conducted.

Theoretically, another approach to relate toxin production

MICROBIOL. REV.

I

j'

'.. ,


http://mm

br.asm.org/

Dow

nloaded from



500 1000 1500V v V

-5 0 10 20 30

ug-t-toxin

200

-150

-100

-50

40 min -5 0 10 20 minFIG. 9. Protective effects of hyperimmune globulin against al-

pha-toxin in vivo. (A) Systolic blood pressure, heart rate (beats per

minute) [bpm]), arterialP02, and blood platelet counts monitored ina cynomolgous monkey that received bolus infusions of alpha-toxinat the depicted times following initial application of 20 ml of saline(V). (B) Similar experiment conducted with a monkey that received4 ml of hyperimmune globulin per kg (V) 5 min prior to applicationof alpha-toxin. Reproduced from Infection and Immunity (20) withpermission.

with clinical relevance is to quantify the level of alpha-toxinin patient serum or body fluids and attempt to make a

correlation with disease. According to one report, alpha-toxin may be present in burn wound infections (55). Thisconclusion was based on the detection of a substance thatproduced lysis of rabbit, but not human, erythrocytes.Unfortunately, methods for unequivocal identification andquantification of the toxin in body fluids or tissues are not yetavailable. In the host organism, alpha-toxin will becomebound to target cells, to antibodies, or to lipoproteins.Reliable quantification methods must therefore first solve thetask of liberating the toxin from these binders. Presently, theonly way to achieve this is to boil samples in SDS (78). Thislowers the sensitivity of available immunoassays to an

extent that renders them unsuitable for diagnostic use. It ishoped that improvement of immunoassays will lead to prog-

ress in this difficult area.

CONCLUSIONS AND PERSPECTIVESAlpha-toxin is doubtless one of the most widespread toxic

agents of bacterial origin encountered by the human organ-

ism. Despite this fact, it has not received the attention itdeserves and the field still poses a number of fascinating andunsolved problems. The characterization of alpha-toxin as a

pore former opened a new field in cell biology by identifyinga major mechanism by which proteinaceous exotoxins can

damage cells. This area is expanding steadily, and pore formersare today recognized as products of many bacterial pathogens.Inasmuch as distinct families of pore-forming toxins have nowbeen identified (e.g., the sulfhydryl-activated toxins with strep-tolysin 0 as a prototype and the family of toxins related to E.coli hemolysin), it may appear surprising the alpha-toxin stillrepresents a distinct entity without recognizable relatedness toany other known pore former (25, 26).How pore formers undergo a hydrophilic-amphiphilic tran-

sition and insert into membranes is not understood at amolecular level, despite the fact that the primary structures ofmany are known. High-resolution structural analysis of theseproteins in both their native and pore-forming states poses anunsolved challenge. Alpha-toxin represents a promising candi-date for such studies as a result of its availability in sufficientlypure form and the relative ease with which it can be handled.No other pore former has been so well characterized with

regard to its primary modes of binding to membranes.Alpha-toxin can interact with specific binding sites on cellsand can absorb in a nonspecific fashion to lipid bilayers.Both types of binding result in the formation of pore-formingoligomers. Neither binding process is understood in molec-ular terms. Are the same or closely associated regions in thetoxin molecule involved in the two types of binding? Whereare they located on the molecule? What is the nature of thehigh-affinity binding site, and on which other cells are thesesites expressed? Could association of the toxin monomerwith the binding site trigger cellular reactions, independentof pore formation, that have hitherto remained unnoticed?The evidence that membrane damage is invoked by toxin

hexamers is compelling. Does hexamerization occur exclu-sively via lateral aggregation? Do oligomers containing fewerprotomers form along with hexamers, and do they inducemembrane damage? Why are certain cells able to tolerate theformation of hexamers? Is the ability to survive due in partto the ability of cells to repair or remove the lesions? If so,what mechanisms underlie these reparative processes, andcould deficiencies exist in disease?

Several pathways via which toxin attack may lead topathological cell reactions have been recognized. Certainly,many others await discovery. Release of interleukin-1 and,possibly, DNA degradation represent but two examples ofsequelae that ad hoc were unsuspected. Does damage uni-formly involve a defined sequence of molecular events, orare different biological functions carried in different regionsof the polypeptide?The potential usefulness of alpha-toxin as a membrane-

permeabilizing agent deserves to be reiterated. The possibilityof producing discrete and stable pores in plasma membranesopens a new avenue for the study of many cellular processesincluding cytokine processing, endonuclease activation, exo-cytosis, enzymatic pathways, and contractile functions.

Finally, medical aspects could gain practical importance.Assuming that alpha-toxin contributes significantly to bacte-rial virulence in humans, could early administration ofhyperimmune globulins be of therapeutic benefit. Will it bepossible to develop sensitive techniques for quantifying thetoxin in body fluids? These queries attest to the fact that agap has now been bridged between basic and applied re-search in the alphatoxin field. More than a century after thediscovery of S. aureus, its major cytolysin has come of age.

ACKNOWLEDGMENTSWork performed on alpha-toxin in the authors' laboratories has

been partially supported by grants from the Deutsche Forschungs-gemeinschaft (project A5, SFB 249, and project D9, SFB 311).We thank Werner Seeger, Norbert Suttorp, Friedrich Grim-

minger, Gudrun Ahnert-Hilger, Ferdinand Hugo, Renate Jursch,and Richard Ward for many stimulating discussions and MargareteHoffmann for excellent secretarial assistance.

REFERENCES1. Abbas-Ali, B., and G. Coleman. 1977. The characteristics of

extracellular protein synthesis by Staphylococcus aureus(Wood 46) and their relationships to the regulation of alpha-toxin. J. Gen. Microbiol. 92:277-282.

A16-

12-

0

4-

1210

CoQ. 8-

42

- 500o -E400-_ <300a't200

a5 100

VOL. 55, 1991

aoImInIn(U *

m z10 0.0 =0z

-614


http://mm

br.asm.org/

Dow

nloaded from



2. Adlam, C., P. D. Ward, A. C. McCartney, J. P. Arbuthnott,and C. M. Thorley. 1977. Effect of immunization with highlypurified alpha- and beta-toxins on staphylococcal mastitis inrabbits. Infect. Immun. 17:250-256.

3. Ahnert-Hilger, G., S. Bhakdi, and M. Gratzl. 1985. Minimalrequirements for exocytosis: a study using PC 12 cells perme-

abilized with staphylococcal a-toxin. J. Biol. Chem. 260:12730-12734.

4. Ahnert-Hilger, G., M. Brautigam, and M. Gratzl. 1987. Ca2+-stimulated catecholamine release from a-toxin-permeabilizedPC12 cells: biochemical evidence for exocytosis and its mod-ulation by protein kinase C and G proteins. Biochemistry26:7842-7848.

5. Arbuthnott, J. P., J. H. Freer, and A. W. Bernheimer. 1967.Physical states of staphylococcal alpha toxin. J. Bacteriol.94:1170-1177.

6. Arbuthnott, J. P., J. H. Freer, and B. Bilicliffe. 1973. Lipid-induced polymerization of staphylococcal alpha toxin. J. Gen.Microbiol. 75:309-319.

7. Arvand, M., S. Bhakdi, B. Dahlback, and K. T. Preissner. 1990.Staphylococcus aureus a-toxin attack on human platelets pro-

motes assembly of the prothrombinase complex. J. Biol.Chem. 265:14377-14381.

8. Bader, M. F., D. Thierse, D. Aunis, G. Ahnert-Hilger, and M.Gratzl. 1986. Characterization of hormone and protein releasefrom a-toxin permeabilized chromaffin cells in primary culture.J. Biol. Chem. 261:5777-5783.

9. Bashford, C. L., G. M. Adler, G. Menestrina, K. J. Micklem,J. J. Murphy, and C. A. Pasternak. 1986. Membrane damageby hemolytic viruses, toxins, complement, and other cytotoxicagents. J. Biol. Chem. 261:9300-9308.

10. Belmonte, G., L. Cescatti, B. Ferrari, T. Nicolussi, M. Ropele,G. Menestrina. 1987. Pore-formation by Staphylococcus au-reus alpha-toxin in lipid bilayers: temperature and toxin con-

centration dependence. Eur. Biophys. J. 14:349-358.11. Bernheimer, A. W. 1974. Interactions between membranes and

cytolytic bacterial toxins. Biochim. Biophys. Acta 344:27-50.12. Bernheimer, A. W., K. S. Kim, C. C. Remsen, J. Antanavage,

and S. W. Watson. 1972. Factors affecting interaction ofstaphylococcal alpha toxin with membranes. Infect. Immun.6:636-642.

13. Bernheimer, A. W., and L. L. Schwartz. 1963. Isolation andcomposition of staphylococcal alpha toxin. J. Gen. Microbiol.30:455-468.

14. Bernheimer, A. W., and L. L. Schwartz. 1965. Effects ofstaphylococcal and other bacterial toxins on platelets in vitro.J. Pathol. Bacteriol. 89:209-223.

15. Bhakdi, S. Unpublished data.16. Bhakdi, S., W. Fassbender, F. Hugo, M.-P. Carreno, C. Ber-

stecher, P. Malasit, and M. D. Kazatchkine. 1988. Relativeinefficiency of terminal complement activation. J. Immunol.141:3117-3122.

17. Bhakdi, S., R. Fussle, and J. Tranum-Jensen. 1981. Staphylo-coccal a-toxin: oligomerisation of hydrophilic monomers to

form amphiphilic hexamers induced through contact with de-oxycholate detergent micelles. Proc. Natl. Acad. Sci. USA78:5475-5479.

18. Bhakdi, S., R. Fussle, G. Utermann, and J. Tranum-Jensen.

1983. Binding and partial inactivation of S. aureus a-toxin by

human plasma low density lipoprotein. J. Biol. Chem. 258:

5899-5904.19. Bhakdi, S., S. Greulich, M. Muhly, B. Eberspacher, H. Becker,

A. Thiele, and F. Hugo. 1989. Potent leukocidal action of

Escherichia coli hemolysin mediated by permeabilization of

target cell membranes. J. Exp. Med. 169:737-754.

20. Bhakdi, S., U. Mannhardt, M. Muhly, F. Hugo, H. Ronne-

berger, and K.-D. Hungerer. 1989. Human hyperimmune glob-ulin protects against the cytotoxic action of staphylococcal

alpha-toxin in vitro and in vivo. Infect. Immun. 57:3214-3220.21. Bhakdi, S., M. Muhly, and R. Ftlssle. 1984. Correlation be-

tween toxin binding and hemolytic activity in membrane dam-

age by staphylococcal a-toxin. Infect. Immun. 46:318-323.

22. Bhakdi, S., M. Muhly, S. Korom, and F. Hugo. 1989. Release

of Interleukin-1l associated with potent cytocidal action ofstaphylococcal alpha-toxin on human monocytes. Infect. Im-mun. 57:3512-3519.

23. Bhakdi, S., M. Muhly, U. Mannhardt, F. Hugo, K. Klapettek,C. MulIer-Eckhardt, and L. Roka. 1988. Staphylococcal a-toxin promotes blood coagulation via attack on human plate-lets. J. Exp. Med. 168:527-542.

24. Bhakdi, S., and J. Tranum-Jensen. 1984. Mechanism of com-plement cytolysis and the concept of channel-forming proteins.Philos. Trans. R. Soc. London Ser. B 306:311-324.

25. Bhakdi, S., and J. Tranum-Jensen. 1987. Damage to mamma-lian cells by proteins that form transmembrane pores. Rev.Physiol. Biochem. Pharmacol. 107:147-223.

26. Bhakdi, S., and J. Tranum-Jensen. 1988. Damage to cellmembranes by pore-forming bacterial cytolysins. Prog. Allergy40:1-43.

27. Bjorklind, A., and S. Arvidson. 1980. Mutants of Staphylococ-cus aureus affected in the regulation of exoprotein synthesis.FEMS Microbiol. Lett. 7:203-206.

28. Blomquist, L., L.-E. Appelgren, and M. Thelestam. 1987.Distribution of 3H-labeled staphylococcal alpha-toxin and atoxin fragment in mice. Infect. Immun. 55:1906-1913.

29. Blomquist, L., T. Bergman, M. Thelestam, and H. Jornvafl.1987. Characterization of domain borders and of a naturallyoccurring major fragment of staphylococcal alpha-toxin. FEBSLett. 211:127-132.

30. Blomquist, L., and A. Sj0gren. 1988. Production and charac-terization of monoclonal antibodies against S. aureus alpha-toxin. Toxicon 26:265-273.

31. Blomquist, L., and M. Thelestam. 1986. Early events in theaction of staphylococcal alpha-toxin on the plasma membraneof adrenocortical Y1 tumor cells. Infect. Immun. 53:636-640.

32. Blomquist, L., and M. Thelestam. 1986. A staphylococcala-toxin fragment: its characterization and use for mappingbiologically-active regions of a-toxin. Acta Pathol. Microbiol.Immunol. Scand. Sect. B 94:277-283.

33. Blomquist, L., and M. Thelestam. 1988. Oligomerization of 3Hlabelled staphylococcal alpha-toxin and fragments on adreno-cortical Y1 tumor cells. Microb. Pathog. 4:223-229.

34. Boehm, D. F., R. A. Welch, and I. S. Snyder. 1990. Domains ofEscherichia coli hemolysin (HlyA) involved in binding ofcalcium and erythrocyte membranes. Infect. Immun. 58:1959-1964.

34a.Bramley, A. J., A. H. Patel, M. O'Reilly, R. Foster, and T. J.Foster. 1989. Roles of alpha-toxin and beta-toxin in virulenceof Staphylococcus aureus for the mouse mammary gland.Infect. Immun. 57:2489-2494.

35. Brown, D. R., and P. A. Pattee. 1980. Identification of achromosomal determinant of alpha-toxin production in Staph-ylococcus aureus. Infect. Immun. 30:36-42.

36. Buckelew, A. R., and G. Colacicco. 1971. Lipid monolayers.Interactions with staphylococcal alpha-toxin. Biochim. Bio-phys. Acta 233:7-16.

37. Burnet, F. M. 1929. The exotoxins of Staphylococcus pyo-genes aureus. J. Pathol. Bacteriol. 32:717-734.

38. Burnet, F. M. 1930. The production of staphylococcal toxin. J.Pathol. Bacteriol. 33:1-16.

39. Cassidy, P., and S. Harshman. 1976. Iodination of a tyrosylresidue in staphylococcal a-toxin. Biochemistry 15:2342-2347.

40. Cassidy, P., and S. Harshman. 1976. Biochemical studies onthe binding of staphylococcal 125 I-labeled alpha-toxin torabbit erythrocytes. Biochemistry 15:2348-2355.

41. Cassidy, P., and S. Harshman. 1979. Characterization of deter-gent-solubilized iodine-125-labeled a-toxin bound to rabbiterythrocytes and mouse diaphragm muscle. Biochemistry 18:232-236.

42. Cassidy, P., H. R. Six, and S. Harshman. 1974. Biologicalproperties of staphylococcal alpha-toxin. Biochim. Biophys.Acta 332:413-423.

43. Cescatti, L., C. Pederzoili, and G. Menestrina. 1991. Modifica-tion of lysine residues of Staphylococcus aureus alpha-toxin:effects on its channel-forming properties. J. Membr. Biol.119:53-64.

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from



44. Chan, K. F., and P. Lazarovici. 1987. Staphylococcus aureusalpha-toxin. 1. Effect on protein phosphorylation in myelin.Toxicon 25:631-636.

45. Clyne, M., J. De Azavedo, E. Carlson, and J. Arbuthnott. 1988.Production of gamma-hemolysin and lack of production ofalpha-hemolysin by Staphylococcus aureus strains associatedwith toxic shock syndrome. J. Clin. Microbiol. 26:535-539.

46. Colacicco, G., and A. R. Buckelew. 1971. Lipid monolayers:influence of lipid film and urea on the surface activity ofstaphylococcal alpha-toxin. Lipids 6:546-553.

47. Coleman, G. 1981. Pleiotropic compensation in the regulationof extracellular protein formation by a low alpha-toxin produc-ing variant of Staphylococcus aureus (Wood 46). J. Gen.Microbiol. 122:11-115.

48. Coleman, G., and B. Abbas-Ali. 1977. Comparison df thepatterns of increase in alpha-toxin and total extracellularprotein by Staphylococcus aureus (Wood 46) grown in mediasupporting widely differing growth characteristics. Infect. Im-mun. 17:278-281.

49. Cooper, L. Z., M. A. Madoff, and L. Weinstein. 1964. Hemo-lysis of rabbit erythrocytes by purified staphylococcal alpha-toxin. J. Bacteriol. 87:136-144.

50. Cooper, L. Z., M. A. Madoff, and L. Weinstein. 1966. Heatstability and species range of purified staphylococcal alpha-toxin. J. Bacteriol. 91:1686-1692.

51. De Christmas, M. J. 1988. Recherches experimentales sur lasuppuration. Ann. Inst. Pasteur 2:469-478.

52. Edelwein, Z., J. Jellaszewicz, S. Smigielski, and C. Zak. 1968.Brain bioelectric activity in staphylococcal a-hemolysin intox-ication. Toxicol. Appl. Pharmacol. 13:133-145.

53. Ericsson, A., M. Granstr0m, R. Mollby, and B. Strandvik.1986. Antibodies to staphylococcal teichoic acid and alpha-toxin in patients with cystic fibrosis. Acta Paediatr. Scand.75:139-144.

54. Fackrell, H. B., and T. E. Hebert. 1987. Inhibition of staphy-lococcal alpha-toxin by covalent modification of an arginineresidue. Biochim. Biophys. Acta 916:419-427.

55. Fader, R. C., P. J. Hals, and F. C. Koo. 1987. Staphylococcaltoxins: screening of burn wound isolates and evidence foralpha-hemolysin production in burn wound. Burns Incl.Therm. Inj. 13:462-468.

56. Fairweather, N., S. Kennedy, T. J. Foster, M. Kehoe, and G.Dougan. 1983. Expression of cloned Staphylococcus aureusalpha-hemolysin determinant in Bacillus subtilis and S. aureus.Infect. Immun. 41:1112-1117.

57. Forlani, L., A. W. Bernheimer, and E. Chiancone. 1971.Ultracentrifugal analysis of staphylococcal alpha toxin. J.Bacteriol. 106:138-142.

58. Forti, S., and G. Menestrina. 1989. Staphylococcal alpha-toxinincreases the permeability of lipid vesicles by cholesterol andpH-depeindent assembly of oligomeric channels. Eur. J. Bio-chem. 181:767-773.

59. Freer, J. H., and J. P. Arbuthnott. 1983. Toxins of Staphylo-coccus aureus. Pharmacol. Ther. 19:55-106.

60. Freer, J. H., J. P. Arbuthnott, and A. W. Bermheimer. 1968.Interaction of staphylococcal alpha toxin with artificial andnatural membranes. J. Bacteriol. 95:1153-1168.

61. Freer, J. H., J. P. Arbuthnott, and B. Billcliffe. 1973. Effects ofstaphylococcal alpha-toxin on the structure of erythrocytemembranes: a biochemical and freeze etching study. J. Gen.Microbiol. 75:321-322.

62. Fussle, R., S. Bhakdi, A. Sziegoleit, J. Tranum-Jensen, T.Kranz, and H. L. Weflensiek. 1981. On the mechanism ofmembrane damage by S. aureus a-toxin. J. Cell Biol. 91:83-94.

63. GemmeDl, C. G., P. K. Peterson, D. J. Schmeling, and P. G.Quie. 1982. Effect of staphylococcal a-toxin on phagocytosis ofstaphylococci by human polymorphonuclear leukocytes. In-fect. Immun. 38:975-980.

64. Gemmell, C. G., P. K. Peterson, K. Townsend, P. G. Quie, andY. Kim. 1982. Biological effects of the interaction of staphylo-coccal a-toxin with human serum. Infect. Immun. 38:981-985.

65. Grant, N. J., D. Aunis, and M. F. Bader. 1987. Morphology andsecretory activity of digitonin and alpha-toxin permeabilized

chromaffin cells. Neuroscience 23:1143-1155.66. Gray, G. S., and M. Kehoe. 1984. Primary sequence of the

a-toxin gene from Staphylococcus aureus Wood 46. Infect.Immun. 46:615-618.

67. Hameed, A., K. J. Olsen, M. Lee, M. G. Lichtenheld, and E. R.Podack. 1989. Cytolysis by Ca-permeable transmembranechannels. Pore formation causes extensive DNA degradationand cell lysis. J. Exp. Med. 169:76-77.

68. Harshman, S., J. E. Alouf, 0. Siffert, and F. Baleux. 1989.Reaction of staphylococcal alpha-toxin with peptide-inducedantibodies. Infect. Immun. 57:3856-3862.

69. Harshman, S., and M. Bondurant. 1985. Susceptibility to sta-phylococcal alpha-toxin of Friend virus-infected murine eryth-roblasts during differentiation. Infect. Immun. 48:114-118.

70. Harshman, S., P. Bouquet, E. Duflot, J. E. Alouf, C. Monte-cucco, and E. Papini. 1989. Staphylococcal alpha-toxin: a studyof membrane penetration and pore formation. J. Biol. Chem.264:14978-14984.

71. Harshman, S., A. M. Burt, J. P. Robinson, M. Blankenship,and D. L. Harshman. 1985. Disruption of myelin sheaths inmouse brain in vitro and in vivo by staphylococcal a-toxin.Toxicon 23:801-806.

72. Harshman, S., and N. Sugg. 1985. Effect of calcium ions onstaphylococcal alpha-toxin-induced hemolysis of rabbit eryth-rocytes. Infect. Immun. 47:37-40.

73. Harshman, S., N. Sugg, B. Gametchu, and R. W. Harrison.1986. Staphylococcal a-toxin: a structure-function study usinga monoclonal antibody. Toxicon 24:403.

74. Haydon, D. A., and S. B. Hladky. 1971. Ion transport acrossthin lipid membranes: a critical discussion of mechanisms inselected systems. Q. Rev. Biophys. 5:187-282.

75. Hildebrand, A., and S. Bhakdi. Unpublishel data.76. Hildebrand, A., M. Roth, and S. Bhakdi. Staphylococcus

aureus alpha-toxin: dual mechanisms of binding to target cells.J. Biol. Chem. 266:17195-17200.

77. Hohman, R. J. 1988. Aggregation of IgE receptor inducesdegranulation in rat basophil leukemia cells permeabilized withalpha-toxin from S. aureus. Proc. NatI. Acad. Sci. USA 85:1624-1628.

78. Hugo, F., A. Sinner, J. Reichwein, and S. Bhakdi. 1987.Quantitation of monomeric and oligomeric forms of mem-brane-bound staphylococcal alpha-toxin by enzyme-linked im-munosorbent assay with a neutralizing monoclonal antibody.Infect. Immun. 55:2933-2939.

79. Ikigai, H., and T. Nakae. 1985. Conformational alteration inalpha-toxin from S.. aureus concomitant with the transforma-tion of the water-soluble monomer to the membrane oligomer.Biochem. Biophys. Res. Commun. 130:175-181.

80. Ikigai, H., and T. Nakae. 1987. Interaction of the alpha-toxin ofS. aureus with the liposome membrane. J. Biol. Chem. 262:2150-2155.

81. Ikigae, H., and T. Nakae. 1987. Assembly of the alpha-toxinhexamer of S. aureus in the liposomal membrane. J. Biol.Chem. 262:2156-2160.

82. Jap, B. K., K. H. Downing, and P. J. Walian. 1990. Structureof PhoE porin in projection at 3.5A resolution. J. Struct. Biol.103:57-63.

83. Jeljaszewicz, J., S. Smiegielski, and W. Hryniewicz. 1978.Biological effects of staphylococcal and streptococcal toxins,p. 185-227. In J. Jeljaszewicz and I. Wadstrom (ed.), Bacterialtoxins and cell membranes. Academic Press, Inc., New York.

84. Jeljaszewicz, J., S. Smigielski, M. Korbecki, and C. Zak. 1965.Histochemical demonstration of changes in enzymatic activityof KB cells produced by staphylococcal alpha and beta hemo-lysins. J. Infect. Dis. 115:421-428.

85. JeUaszewicz, J., S. Szmiegielski, and C. Zak. 1969. Distributionof 1311-labelled staphylococcal alpha-hemolysin in the rabbit.Zentralbi. Bakteriol. Parasitenkd. Infektionskr. Hyg. ErsteAbt. Orig. Reihe B 209:310-314.

86. Jonsson, P., M. Lindberg, I. Haraldsson, and T. Wadstrom.1985. Virulence of staphylococcus aureus in a mouse mastitismodel: studies of alpha hemolysin, coagulase, and protein A as

possible virulence determinants with protoplast fusion and

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from



gene cloning. Infect. Immun. 49:765-769.87. Julander, I. G., M. Grandstr0m, S. A. Hedstr0m, and R.

M0olby. 1983. The role of antibodies against alpha-toxin andteichoic acid in the diagnosis of staphylococcal infections.Infection 11:77-83.

88. Kapral, F. A., J. R. Godwin, and S. E. Dye. 1980. Formation ofintraperitoneal abscesses by Staphylococcus aureus. Infect.Immun. 30:204-211.

89. Kato, I. 1979. Inhibitory effect of a lethal toxic fragment ofstaphylococcal a-toxin on cyclic AMP-dependent protein ki-nase activity. Biochim. Biophys. Acta 570:388-396.

90. Kato, I., and M. Naiki. 1976. Ganglioside andrabbit erythro-cyte membrane receptor of staphylococcal alpha-toxin. Infect.Immun. 13:289-291.

91. Kato, I., K. Sakoda, M. Saito, Y. Suzuki, and M. Watanabe.1975. Inhibitory effect of flavin mononucleotide on the hemo-lysis of rabbit erythrocytes by staphylococcal alpha-toxin.Infect. Immun. 12:696-697.

92. Kato, I., K. Sakoda, M. Saito, and Y. Suzuki. 1977. Effects oflectins on the hemolysis of rabbit erythrocytes by staphylococ-cal alpha-toxin. Microbiol. Immunol. 21:517-524.

93. Kato, I., and M. Watanabe. 1980. Chemical studies on staphy-lococcal alpha toxin and its fragments. Toxicon 18:361-365.

94. Kato, I., H. Watanabe, and N. H. Kumazawa. 1979. Inhibitionof adenosine-3',5'-monophosphate-dependent protein kinaseby staphylococcal alpha-toxin. Infect. Immun. 24:286-288.

94a.Kehoe, M. Personal communication.95. Kehoe, M., I. Duncan, T. Foster, N. Fairweather, and G.

Dougan. 1983. Cloning, expression, and mapping of the Staph-ylococcus aureus alpha-hemolysin determinant in Escherichiacoli K-12. Infect. Immun. 41:1105-1111.

96. Kinsky, S. C. 1970. Antibiotic interaction with model mem-

branes. Annu. Rev. Pharmacol. 10:119-142.97. Koch, R. 1878. Untersuchungenuber die Aetiologie der Wund-

infektionskrankheiten. Ges. Werke 1:61-108.98. Kumar, S., K. Locken, A. Kenyon, and R. K. Lindorfer. 1962.

The characterization of staphylococcal toxins. II. The isolationand characterization of a homogeneous staphylococcal proteinpossessing alpha hemolytic, dermonecrotic, lethal and leuco-cidal activities. J. Exp. Med. 115:1107-1115.

99. Lazarovici, P., and K. F. Chan. 1987. S. aureus alpha-toxin. 2.

Reduction of epidermal growth factor receptor affinity in PC12cells. Toxicon 25:637-647.

100. Lee, K. Y., and T. H. Birkbeck. 1985. Effect of phenethylalcohol on Staphylococcus aureus alpha-lysin production. In-

fect. Immun. 7:112-117.101. Lind, I., G. Ahnert-Hilger, G. Fuchs, and M. Gratzl. 1987.

Purification of alpha-toxin from S. aureus and application to

cell permeabilization studies. Anal. Biochem. 164:84-89.102. Lominski, L., J.-P. Arbuthnott, and J. B. Spence. 1963. Purifi-

cation of staphylococcus alpha-toxin. J. Pathol. Bacteriol.

86:258-262.103. Ludwig, A., M. Vogel, and W. Goebel. 1987. Mutations af-

fecting activity and transport of heamolysin in Escherichia

coli. Mol. Gen. Genet. 206:238-245.104. Maharaj, I., and H. B. Fackrell. 1980. Rabbit erythrocyte band

3: a receptor for staphylococcal alpha toxin. Can. J. Microbiol.

26:524.105. Manohar, M., S. K. Maheswaran, S. P. Frommes, and R. K.

Lindorfer. 1967. Platelet damaging factor, a fifth activity of

staphylococcal a-toxin. J. Bacteriol. 94:224-231.106. Mayer, M. M. 1972. Mechanism of cytolysis by complement.

Proc. Natl. Acad. Sci. USA 69:2954-2959.107. Mayorek, N., and J. Bar-Tana. 1989. Lipid synthesis in per-

meabilized cultured rat hepatocytes. J. Biol. Chem. 264:4450-4455.

108. McCartney, C. A., and J. P. Arbuthnott. 1978. Mode of action

of membrane damaging toxins produced by staphylococci, p.

89-127. In T. Wadstrom and J. Jeljaszewicz (ed.), Bacterial

toxins and cell membranes. Academic Press Inc., London.

109. McConkey, D. J., S. Orrenius, and M. Jondal. 1990. Cellular

signalling in programmed cell death (apoptosis). Immunol.

Today 11:120-121.

110. McEwen, B. F., and W. J. Arion. 1985. Permeabilization of rathepatocytes with staphylococcus aureus a-toxin. J. Cell Biol.100:1922-1929.

111. McGee, M. P., A. Kreger, E. S.Leake,and S. Harshman. 1983.Toxicity of staphylococcal alpha toxin for rabbit alveolarmacrophages. Infect. Immun. 39:439.

112. McNiven, A. C., P. Owen, and J. P. Arbuthnott. 1972. Multipleforms of staphylococcal alpha-toxin. J. Med. Microbiol. 5:113-122.

113. Menestrina, G. Personal communication.114. Menestrina, G. 1986. Ionic channels formed by staphylococcus

aureus a-toxin: voltage-dependent inhibition by divalent andtrivalent cations. J. Membr. Biol. 90:177-190.

115. Morfeldt, E., L. Janzon, S. Arvidson, and S. Lofdahl. 1988.Cloning of a chromosomal locus (exp) which regulates theexpression of several exoprotein genes in Staphylococcusaureus. Mol. Gen. Genet. 211:435-440.

116. Nishlmura, J., M. Kolber, and C. van Breemen. 1988. Norepi-nephrine and GTP-gamma-S increase myofilament Ca2'-sensi-tivity in alpha-toxin permeabilized arterial smooth musclecells. Biochem. Biophys. Res. Commun. 157:677-683.

117. Ogston, A. 1881. Report upon microorganisms in surgicaldiseases. Br. Med. J. May:369-375.

118. Olofsson, A., U. Kaveus, I. Hackseil, M. Thelestam, and H.Hebert. 1990. Crystalline layers and three-dimensional structureof staphylococcus aureus a-toxin. J. Mol. Biol. 214:299-306.

119. Olofsson, A., U. Kaveus, M. Thelestam, and H. Hebert. 1988.The projection structure of alpha-toxin from S. aureus inhuman platelet membranes as analysed by electron microscopyand image processing. J. Ultrastruct. Mol. Struct. Res. 100:194-200.

120. O'Reilly, M., J. C. S. Azavedo, S. Kennedy, and T. J. Foster.1986. Inactivation of the alpha-haemolysin gene of staphylo-coccus aureus 8325-4 by site-directed mutagenesis and studieson the expression of its haemolysins. Microb. Pathog. 1:125-138.

120a.O'Reilly, M., B. Kreiswirth, and T. J. Foster. 1990. Crypticalpha-toxin gene in toxic shock syndrome and septicaemiastrains of Staphylococcus aureus. Mol. Microbiol. 4:1947-1955.

121. Palmer, M., and S. Bhakdi. Unpublished data.122. Patel, A. H., P. Nowlan, E. D. Weavers, and T. Foster. 1987.

Virulence of protein A-deficient and alpha-toxin-deficient mu-tants of Staphylococcus aureus isolated by allele replacement.Infect. Immun. 55:3103-3110.

123. Pattee, P. A. 1986. Chromosomal map location of the alpha-hemolysin structural gene in Staphylococcus aureus NCTC8325. Infect. Immun. 54:593-5%.

124. Pederzoili, C., L. Cescatti, and G. Menestrina. 1991. Chemicalmodification of Staphylococcus aureus alpha-toxin by diethyl-pyrocarbonate: role of histidines in its membrane-damagingproperties. J. Membr. Biol. 119:41-52.

125. Petrini, B., and R. Moilby. 1981. Activation of human lympho-cytes in vitro by membrane-damaging toxins from Staphylo-coccus aureus. Infect. Immun. 31:592-956.

126. Phimister, G. M., and J. H. Freer. 1984. Binding of _251-alphatoxin of staphylococcus aureus to erythrocytes. J. Med. Mi-crobiol. 18:197-204.

127. Raff, M. J., P. Barnwell, and A. S. Werner. 1977. Detoxifica-tion of staphylococcal alpha-toxin by hydrocortisone andmethylprednisolone. J. Med. Microbiol. 11:67-73.

128. Recsei, P., B. Kreiswirth, M. O'Reilly, P. Schlevat, A. Gruss,and R. V. Novick. 1986. Regulation of exoprotein gene expres-sion in Staphylococcus aureus by agr. Mol. Gen. Genet.202:58-61.

129. Reichwein, J., F. Hugo, M. Roth, A. Sinner, and S. Bhakdi.1987. Quantitative analysis of the binding and oligomerizationof staphylococcal alpha-toxin in target erythrocyte mem-branes. Infect. Immun. 55:2940-2944.

130. Rogolsky, M. 1979. Nonenteric toxins of Staphylococcus au-reus. Microbiol. Rev. 43:320-360.

131. Schneider, E., C. W. M. Haest, G. Plasa, and B. Deuticke. 1986.Bacterial cytotoxins, amphotericin B and local anesthetics

MICROBIOL. REV.


http://mm

br.asm.org/

Dow

nloaded from



enhance transbilayer mobility of phospholipids in erythrocytemembranes. Consequences for phospholipid asymmetry. Bio-chim. Biophys. Acta 855:325-336.

132. Schrezenmeier, H., G. Ahnert-Hilger, and B. Fleischer. 1988.Inactivation of a T cell receptor-associated GTP-binding pro-tein by antibody-induced modulation of the T cell receptor/CD3 complex. J. Exp. Med. 168:817-822.

133. Seeger, W., M. Bauer, and S. Bhakdi. 1984. Staphylococcala-toxin elicits hypertension in isolated rabbit lungs due tostimulation of the arachidonic acid cascade. J. Clin. Invest.74:849-858.

134. Seeger, W., R. G. Birkemeyer, L. Ermert, N. Suttorp, S.Bhakdi, and H. R. Duncker. 1990. Staphylococcal alpha-toxininduced vascular leakage in isolated perfused rabbit lungs.Lab. Invest. 63:341-349.

135. Seydel, U., G. Schroder, and K. Brandenburg. 1989. Reconsti-tution of the lipid matrix of the outer membrane of Gram-negative bacteria as asymmetric planar bilayer. J. Membr.Biol'. 109:95-103.

136. Siegel, I., and S. Cohen. 1964. Action of staphylococcal toxinon human platelets. J. Infect. Dis. 114:488-502.

137. Six, H., and S. Harshman. 1973. Purification and properties oftwo forms of staphylococcal a-toxin. Biochemistry 12:2672-2677.

138. Six, H., and S. Harshman. 1973. Physical and chemical studieson staphylococcal a-toxins A and B. Biochemistry 12:2677-2683.

139. Smigielski, S., M. Blankenship, J. P. Robinson, and S. Harsh-man. 1979. Injury of myelin sheaths in isolated rabbit vagusnerves by a-toxin of staphylococcus aureus. Toxicon 17:363-371.

140. Suttorp, N. Unpublished data.141. Suttorp, N., and E. Habben. 1988. Effect of staphylococcal

alpha-toxin on intracellular Ca2+ in polymorphonuclear leuko-cytes. Infect. Immun. 56:2228-22354.

142. Suttorp, N., T. Hessz, W.'Seeger, A. Wilke, R. Koob, F. Lutz,and D. Drenckhahn. 1988. Bacterial exotoxins and endothelialpermeability for water and albumin in vitro. Am. J. Physiol.255:C368-C377.

143. Suttorp, N., W. Seeger, E. Dewein, S. Bhakdi, and L. Roka.1985. Staphylococcal a-toxin stimulates synthesis of prostacy-clin by cultured endothelial cells from pig pulmonary arteries.Am. J. Physiol. 248:C127-C135.

144. Suttorp, N., W. Seeger, J. Zucker-Reimann, and S. Bhakdi.1987. Mechanism of leukotriene generation of polymorphonu-clear leukocytes by staphylococcal alpha-toxin. Infect. Im-mun. 55:104-110.

145. Taubler, J. H., F. A. Kapral, and S. Mudd. 1963. Role ofalpha-toxin in lesion formation by Staphylococcus aureus onsutures subcutaneously implanted in mice. J. Bacteriol. 86:51-57.

146. Thai, A., and W. Egner. 1954. Local effect of staphylococcaltoxin. Studies on blood vessels with particular reference tophenomenon of dermonecrosis. Arch. Pathol. 57:392-404.

147. Thai, A., and W. Egner. 1961. Site of location of staphylococ-cus alpha toxin. J. Exp. Med. 113:67-82.

148. Thelestam, M., and L. Blomquist. 1988. Staphylococal alpha-toxin-recent advances. Toxicon 26:51-65.

149. Thelestam, M.,'C. Jolivet-Reynaud, and J. E. Alouf. 1983.Photolabeling of staphylococcal a-toxin from within rabbiterythrocyte membranes. Biochem. Biophys. Res. Commun.111:44449.

150. Thelestam, M., and R. Moliby. 1975. Determination of toxin-induced leakage of different-size nucleotides through theplasma membranes of human diploid fibroblasts. Infect. Im-mun. 11:640-648.

151. Thelestam, M., and R. MoUby. 1975. Sensitive assay fordetection of toxin-induced damage to the cytoplasmic mem-brane of human diploid fibroblasts. Infect.Immun. 12:225-232.

152. Thelestam, M., and R. MoUby. 1983. Survival of cultured cellsafter functional and structural disorganization of plasma mem-brane by bacterial hemolysins and phospholipases. Toxicon21:805-815.

153. Thelestam, M., R. MoDlby, and T. Wadstrom. 1973. Effects ofstaphylococcal alpha-, beta-, delta-, and gamma-hemolysins onhuman diploid fibroblasts and HeLa cells. Evaluation of a newquantitative assay for measuring cell damage. Infect. Immun.8:938-946.

154. Thielens, N. M., K. Lohner, and A. F. Esser. 1988. Humancomplement protein C9 is a calcium binding protein. Structuraland functional implications. J. Biol. Chem. 263:6665-6670.

155. Tobkes, N., B. A. Wallace, and H. Bayley. 1985. Secondarystructure and assembly mechanism of an oligomeric channelprotein. Biochemistry 24:1915-1920.

156. Tranum-Jensen, J., and S. Bhakdi. 1987. The action of com-plement and channel-forming bacterial toxins on cell mem-branes, p. 445-492. In J. R. Harris and R. W. Home (ed.),Electron microscopy of proteins, vol. 6. Academic Press Inc.,London.

157. Tweten, R. K., K. K. Christianson, and J. J. landolo. 1983.Transport and processing of staphylococcal alpha-toxin. J.Bacteriol. 156:524-528.

158. Van der Merwe, P., R. P. Millar, I. K. Wakefield, and J. S.Davidson. 1989. Mechanisms of luteinizing-hormone exocyto-sis in staphylococcus aureus a-toxin permeabilized sheepgonadotropes. Biochem. J. 264:901-908.

159. Van der Vijver, J. C. M., M. M. van Es-Boom, and M. F.Michel. 1975. A study of virulence factors with induced mu-tants of Staphylococcus aureus. J. Med. Microbiol. 8:279-287.

160. Vann, J. M., and C. A. Proctor. 1988. Cytotoxic effects ofingested S. aureus on bovine endothelial cells: role of S. aureusalpha-haemolysin. Microb. Pathog. 4:443-453.

161. Vanstapel, F., and N. Blanckaert. 1987. Endogenous esterifica-tion of bilirubin by liver microsomes. J. Biol. Chem. 262:4616-4623.

161a.von Leber, T. 1888. Uber die Entstehung der Entzundung unddie Wirkung der entzundungserregenden Schadlichkeiten.Fortschr. Med. 6:460-464.

162. Ward, P. D., C. Adlam, A. C. McCartney, J. P. Arbuthnott,and C. M. Thorley. 1979. A histopathological study of theeffects of highly purified staphylococcal alpha- and beta-toxinson the lactating mammary glands and skin of the rabbit. J.Comp. Pathol. 89:169-177.

163. Watanabe, M., and I. Kato. 1978. Purification and someproperties of a lethal toxic fragment of staphylococcal a-toxinby tryptic digestion. Biochim. Biophys. Acta 535:388-400.

164. Watanabe, W., T. Tomita, and T. Yasuda. 1987. Membrane-damaging action of staphylococcal alpha-toxin on phospholip-id-cholesterol liposomes. Biochim. Biophys. Acta 898:257-265.

165. Weissman, G., G. Sessa, and A. W. Bernheimer. 1966. Staphy-lococcal a-toxin: effects on artificial lipid spherules. Science154:772-774.

166. Wiedmer, T., B. Ando, and P. J. Sims. 1987. ComplementCSb-9 stimulated platelet secretion is associated with a Ca2+-initiated activation of cellular protein kinases. J. Biol. Chem.262:13681-13686.

167. Wiedmer, T., C. R. Esmon, and P. J. Sims. 1986. On themechanism by which complement proteins C5b-9 increaseplatelet prothrombinase activity. J. Biol. Chem. 261:14587-14592.

168. Wiseman, G. M. 1975. The hemolysins of Staphylococcusaureus. Bacteriol. Rev. 39:317-344.

169. Wiseman, G. M., and J. D. Caird. 1972. Further observationson the mode of action of the alpha toxin of Staphylococcusaureus "Wood-46". Can. J. Microbiol. 18:987-992.

170. Wiseman, G. M., J. D. Caird, and H. B. Fackrell. 1975.Trypsin-mediated activation of the alpha haemolysin of Staph-ylococcus aureus "Wood 46". J. Med. Microbiol. 8:29-38.

171. Wurzel, M., A. W. Bernheimer,and B. W. Zweifach. 1966.Staphylococcal alpha toxin: induced contraction of isolatedarterial smooth muscle. Am. J. Physiol. 210:360-364.

172. Yoshihawa, M., F. Matsudo, M. Naka, E. Murafushi, and Y.Tsunemetsu. 1974. Pleiotropic alteration of activities of severaltoxins and enzymes in mutants of Staphylococcus aureus. J.Bacteriol. 119:143-148.

VOL. 55, 1991


http://mm

br.asm.org/

Dow

nloaded from


alpha-toxin of staphylococcus aureuss. aureus alpha-toxin 735 at sedimentation equilibrium was...

Documents