multiple oligomeric states of the helicobacter pylori vacuolating toxin demonstrated by...

Multiple Oligomeric States of the Helicobacter pyloriVacuolating Toxin Demonstrated by Cryo-electron Microscopy

Marc Adrian1, Timothy L. Cover2,3*, Jacques Dubochet1 andJohn E. Heuser4

1Laboratoire d’AnalyseUltrastructurale, Batiment deBiologie, Universite deLausanne, LausanneSwitzerland

2Departments of Medicine andMicrobiology and ImmunologyVanderbilt University School ofMedicine, Nashville, TN 37232USA

3Department of VeteransAffairs Medical CenterNashville, TN 37212, USA

4Department of Cell BiologyWashington University Schoolof Medicine, St. Louis, MO63110, USA

Helicobacter pylori vacuolating toxin (VacA) is a bacterial protein toxin thatforms water-soluble oligomeric complexes, and can somehow insert intolipid bilayers to produce anion-selective channels. In this study, we utilizethe novel technique of “cryo-negative staining” to examine themorphology of vitrified VacA complexes. Two basic types of oligomericstructures were observed: (i) relatively thick six or seven-sided astralarrays with near-perfect radial symmetry; and (ii) relatively thin astralarrays of six to nine short “rodlets” that display a distinct handedness or“chirality”. Additionally, the new technique provided edge-views of thethicker form of VacA oligomer, which appears to be a thin bilayered disc,indicating that the relatively thick six-sided arrays are actuallydodecamers. Also observed occasionally in the present cryo-negativelystained VacA preparations were 2D crystalline arrays that appeared to becomprised of interlocked dodecamers. The structural alterations thatVacA oligomers must undergo to form these 2D crystals were analyzed,and intermediates in this transition were identified. Additionally, the oli-gomeric state of acid-activated VacA bound to membranes was visualizedby the traditional technique of “deep-etch” electron microscopy, and wasfound to resemble most closely the top halves of the dodecamers. Theseresults indicate that VacA is able to undergo major conformationalchanges, accompanied by major changes in its state of oligomerization,under different natural and experimental conditions.

q 2002 Elsevier Science Ltd. All rights reserved

Keywords: molecular structure; macromolecular assembly; ion channel;gastric cancer; peptic ulcer*Corresponding author

Introduction

Helicobacter pylori are Gram-negative bacteriathat colonize the stomachs of more than half theworld’s human population. Colonization of thegastric mucosa by H. pylori consistently producesan inflammatory response, termed chronic super-ficial gastritis. The presence of H. pylori in the gas-tric mucosa is an important risk factor for thedevelopment of peptic ulcer disease, gastric adeno-carcinoma, and gastric lymphoma.1

Of the many bacterial species that are ingestedeach day along with food, only H. pylori and a

related species (H. heilmannii ) are able to persist-ently colonize the human stomach. Adaptation ofH. pylori for life in this hostile acidic environmenthas been found to be associated with the synthesisof various novel bacterial products that can operatein such a milieu. One such factor is a secreted toxin(VacA) that is unrelated to any other knownbacterial protein toxin (see Refs. 2,3 for reviews).H. pylori VacA produces several different effectson eukaryotic cells. The most striking is thedevelopment of large cytoplasmic vacuoles thatpossess both late endosomal and lysosomalmarkers in their limiting membranes.4 VacA-treated cells exhibit defects in the maturation andtrafficking of lysosomal hydrolases.5 In addition,VacA interferes with the process of antigenpresentation,6 and induces increased permeabilityacross epithelial cell monolayers.7 The mechanisms

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved

E-mail address of the corresponding author:[email protected]

Abbreviations used: EM, electron microscopy; AFM,atomic force microscopy.

doi: 10.1016/S0022-2836(02)00047-5 available online at http://www.idealibrary.com onBw

J. Mol. Biol. (2002) 318, 121–133

underlying these many different VacA actions arenot understood completely, but are thought toinvolve disruption in normal membrane traffickingwithin the endosomal pathway3,4,8 and/or theformation of anion-selective channels in the plasmamembrane and in endosomal membranes.9 – 13

Several lines of evidence suggest that VacA playsan important role in the pathogenesis of H. pylori-associated illnesses in humans, including pepticulcer disease.2,3

H. pylori VacA is synthesized initially as a140 kDa precursor protein that undergoes amino-terminal and carboxy-terminal proteolytic proces-sing to yield a mature 88 kDa secreted protein.14 – 17

The secreted 88 kDa VacA monomers then self-

assemble into large water-soluble complexes witha molecular mass of ,900 kDa.14 These complexeshave been visualized by deep-etch electronmicroscopy.18 – 20 Figure 1 (rows 1–3) shows an“anaglyph” stereo-view of the three basic types ofimages of VacA that have been identified afterdeep-etching a preparation of ,900 kDa oligomersadsorbed to mica flakes. The top row of Figure 1shows the most frequently observed view of VacAadsorbed to mica: a relatively thick, 30 nmdiameter, “flower” or “snowflake” shaped entitycomposed of a ,15 nm central ring surroundedby six or seven symmetrically arranged “petals”.In fact, sophisticated 3D reconstructions of deep-etchings done by others have shown that the petalsof this “thick-form” of VacA oligomer display aslight counterclockwise cant or skew.18,20 However,despite this slight asymmetry, all thick-forms seenafter deep-etching look exactly the same,suggesting that they must be shaped the same onboth sides. (This we conclude with some certainty,because we see no evidence that VacA is not freeto adsorb to mica by either side, and thus to exposeboth sides to replication and viewing.)

The two less frequently observed views of VacAseen after deep-etching look relatively thin (e.g.they are flatter or less raised from the mica sub-strate). These appear to consist of six or seven6 nm £ 14 nm bent “rodlets” that radiate obliquelyout from the center of the complex with a distinctcant or skew. Due to their intrinsic chirality, thesethin-forms display two different, roughly mirror-symmetric images, depending upon which sideadsorbs to the mica and, hence, which side faces“up” and is imaged after deep-etching. When seenin the clockwise orientation, they look like six orseven relatively straight rodlets in a simple astralarray (Figure 1, row 3); whereas in counterclock-wise views (which are found very infrequently, forreasons discussed below) they appear as an arrayof six or seven teardrop-shaped subunits withtheir bulbous domains oriented distally (Figure 1,row 2). Anaglyph stereo viewing of these counter-clockwise views demonstrates that they lookalmost exactly the same as the thicker formsshown in Figure 1, row 1. In particular, they dis-play the same central ,15 nm ring. The onlydifference is that they are clearly thinner (e.g. flat-ter or less raised from the mica substrate).

We have proposed a model of VacA constructionshown in Figure 2.19 The thick, relatively symmetri-cal flower form of VacA illustrated at the bottom ofthis Figure is drawn as if it is a dodecamer formedby face-to-face apposition of the two thinnerhexamers drawn above. The upper right drawingpresents the rarely observed thin hexamer of VacAthat shows counterclockwise chirality and looksalmost exactly like the thicker form, even down toits central ring, because it presents the same viewas the thicker form. The upper left drawingpresents the more commonly observed clockwiseview of the thin hexamer of VacA. It looks entirelydifferent from the thicker form because, according

Figure 1. 3D “anaglyph” view of VacA adsorbed tomica flakes and imaged by deep-etch EM, showing thethree different sorts of views obtained. Row 1: relativelythick, radially symmetric flowers exhibiting a prominentcentral ring. Row 2: flatter “thin-forms” with counter-clockwise chirality and a visible central ring. Row 3:equally flat thin-forms with clockwise chirality that lacka prominent central ring. (See Figure 2 for an inter-pretation of these three views.) Rows 4 and 5: intactVacA oligomers adsorbed to mica were treated with pH3.5 glycine buffer. Note that they dissociate into 12–14distinct “rodlike” components. Magnification, 300,000 £ .

122 Oligomeric States of H. pylori Vacuolating Toxin

to the model, it presents the surface of the hexamerthat is normally “sandwiched” in the interior ofthe dodecamer, and hence normally inaccessible toview. (This would presumably be its more thehydrophobic surface, which, due to its hydro-phobicity, would have a much lower probabilityof adhering to mica, thereby explaining the rarityof its landing “face down”, as must occur for thehexamer to present its opposite, counterclockwiseview). Figure 1 (rows 4 and 5) provides criticalsupport for this model, by showing that when apreparation of VacA oligomers that have been pre-adsorbed to mica is subsequently exposed to lowpH (a treatment that is known to enhance itscytotoxicity),19,21 – 24 the thicker flower forms ofVacA are converted quantitatively into astrallyarrayed clusters of 10–14 individually distinguish-able rodlets that look exactly like those seen in theclockwise interior views of VacA (Figure 1, row 3),and exactly like isolated 88 kDa VacA monomers.19

In addition to studies of VacA structure by deep-etch electron microscopy (EM), atomic forcemicroscopy (AFM) has been used to analyze thestructure of VacA bound to lipid bilayers restingon solid substrates.9 To date, AFM has not beenable to resolve individual VacA oligomers, but ithas been able to image 2D crystals when VacAwas added to lipid substrates as acidified mono-mers and then re-neutralized to promote theirreassociation.9 The subunit spacing of these 2Dcrystals suggested that the VacA oligomers hadachieved a considerable overlap or interdigitation.The unit cells in these 2D crystals displayed promi-nent central rings that appeared similar in struc-

ture to the rings in the centers of the individualVacA flowers seen by deep-etch EM. However, theAFM seemed to indicate that the 2D crystals weretoo thin to be composed of two-layered dodeca-meric structures. Thus, it was proposed that thecrystals represent a single layer of interlocked,symmetrical, hexameric “thin-forms”.9

One difficulty in interpreting these various deep-etch EM and AFM images is that with both tech-niques only the exposed surfaces of VacA can bevisualized (i.e. the surfaces that face away fromthe supporting substrate of mica or lipid). Thus,several different possible interpretations andmodels of VacA structure have beenproposed.9,18 – 20,25 In an effort to overcome thelimitations of previous deep-etch EM and AFMimaging methods, and to gain further insight intoVacA structure, we utilize in the present study therelatively new technique of cryo-negative staining.This involves the vitrification of thin layers ofwater-soluble VacA oligomers suspended in aconcentrated ammonium molybdate solution, fol-lowed by high-resolution cryo-EM visualization.26

This approach yields images that are generallysuperior to those obtained by conventional nega-tive staining, simply because the latter approachnormally involves a harsh air-drying step. Indeed,this new approach has allowed us to visualize anassortment of different oligomeric forms of VacA.These include the familiar thick, symmetrical com-plexes and the “thinner” chiral forms, as well asthe 2D crystalline arrays seen by AFM. Addition-ally it has provided images of side views of thesecomplexes not available with any of the other

Figure 2. The model of VacAoligomeric structure we proposedin a previous study19 and furtheradvocate here. Bottom: the thicksymmetrical “flower”-form of VacAis interpreted to be a dodecamerformed by face-to-face appositionof the two hexamers shown earlier.Upper right: the thin-form withcounterclockwise chirality repre-sents the top half of the dodecamer.Upper left: the thin form withclockwise chirality represents thebottom half of the dodecamer, withits hydrophobic surface facing up.(According to the model, this facewould normally be sandwichedinside the dodecamer and wouldprovide the “glue” for holding thesandwich together.)

Oligomeric States of H. pylori Vacuolating Toxin 123

preparative techniques. These, in particular, haveallowed us to reconcile the multiple existinginterpretations of VacA’s oligomeric structure andto validate our original model for how VacAoligomerizes in an aqueous versus a lipidicenvironment.

Results

Cryo-negative staining of isolatedVacA oligomers

Analysis of preparations of purified VacA bycryo-negative staining revealed a mixture ofseveral different ,25 nm diameter flower or snow-flake-shaped molecules, including well-contrastedsymmetrical structures and less-contrasted chiralstructures (Figure 3). Closer examination of thesemolecules demonstrated that the well-contrastedsymmetrical structures were composed of a central,13 nm ring surrounded by six or seven 3 nm £7 nm peripheral “spikes” (Figure 4). The secondclass of molecules appeared to consist of astralclusters of thin 3 £ 13 rods that radiated outwardswith a distinct cant or chirality (Figure 5). Thesechiral, less-contrasted oligomers displayed a vari-able number of rod-like projections, ranging from

six to nine. Interestingly, the handedness of thesechiral forms was always the same: namely,counterclockwise. This indicated that they werenot actually floating in the stain, but wereadsorbed to one (and only one) of its air/waterinterfaces. (See Ref. 27 for an explanation of howthis phenomenon occurs, and why only onesurface of the preparation usually “captures” themolecules.) Analyses of several different prepa-rations of VacA demonstrated that these relativelyfaint, chiral oligomers were typically much moreabundant than the better-contrasted symmetricalforms (Figure 3). Additionally, large numbers ofindividual ,13 nm rodlets (presumably represent-ing VacA monomers) were seen in the backgroundof all the cryo-negatively stained preparations(Figure 3).

Besides the two distinctive en face views of VacAoligomer shown in Figures 4 and 5, seen also inthe thicker regions of cryo-negatively stainedpreparations, were ,25 nm wide, plate-like struc-tures composed of two closely opposed flat discs(Figure 6). The central regions of these complexesappeared to be attached to each other quite closely,whereas the peripheral “spikes” did not. The thick-ness of these forms was ,9 nm. In Discussion, weargue that these represent side-views of the thicker,symmetrical VacA complexes shown in Figure 4.

Figure 3. Low-magnificationoverview of a cryo-negativelystained preparation of VacA puri-fied from broth-culture supernatantof H. pylori strain 60190. Severaldifferent types of oligomer arepresent, most prominently therelatively distinct and symmetrical“snowflakes”, all against a faintbackground of less well contrastedchiral oligomers and monomers.Magnification, 10,000 £ .


No side-views of the thinner, chiral forms wereobserved (for the obvious reason that they representcomplexes adsorbed to the air–water interface).

Comparison of deep-etch EM and cryo-negatively stained images of VacA

One striking difference between the cryo-negatively stained VacA images presented here

and previous deep-etch imagesis that the thinner, chiral forms predominated inthe present preparations (Figure 3), whereas thethicker, symmetrical forms predominated in earlierdeep-etch images.18,19 In addition, the “chiral”forms never displayed as many as eight or nineperipheral spikes in earlier deep-etchpreparations,18,19 but such forms were relativelyabundant in the present cryo-negatively stained

Figure 4. Cryo-negative stainedimages of the symmetrical, well-contrasted flower-form of the VacAcomplex. A prominent central ringis visible within each of these struc-tures. Six or seven visible spikesradiate symmetrically outwardfrom the center of each complex. Ina few cases (cf. row 4, leftmostpanel) these spikes display acounterclockwise “cant” andappear to possess slight hooks attheir ends. The lower leftmostpanel displays, at the bottom, asecond type of complex shown inmore detail in Figure 5. Magnifi-cation, 300,000 £ .

Figure 5. Cryo-negatively stainedimages of the relatively thin chiralforms of VacA complexes. In con-trast to those shown in Figure 4,these structures were consistentlyless well contrasted and their armsradiated from the center with a dis-tinct counterclockwise chirality orskew. In addition, there was con-siderable variability in the numberof arms, ranging from 6 to 9.The bottom right panels displayoligomeric complexes that havedissociated into individual, rod-like components. Magnification,300,000 £ .


preparations. To exclude the possibility that thesedifferences might have resulted from systematicdifferences in the methods for culturing H. pyloriand purifying VacA, two different aliquots of oneparticular preparation of VacA were imaged byboth techniques simultaneously. The deep-etchedversion of this preparation again showed a prepon-derance of symmetrical six-sided forms with rela-tively few chiral thin-forms, and the latter werecomposed of no more than six or seven interlockedrodlets. In contrast, the aliquot examined by cryo-

negative staining again looked just as describedabove; namely, it displayed a vast preponderanceof individual rodlets, among which were founda large number of seven–nine-membered chiralthin-forms, but relatively few of the thicker, sym-metrical six-sided forms.

A second difference between previous deep-etchimages of VacA and the present cryo-negativelystained images is that the peripheral petals of thecomplexes appeared considerably narrower in thepresent images. Thus, we described them above as

Figure 6. Cryo-negatively stainedimages of bilayered discs. The cen-tral regions of these ,25 nm widediscs appear to form closelyopposed, stain-excluding sand-wiches, some ,9 nm thick, whereastheir radiating edges do not appearto be in direct contact. These areinterpreted to be side-views of thecomplexes shown in Figure 4. Noexamples of side-views of the chiralforms shown in Figure 5 werefound. Magnification, 300,000 £ .

Figure 7. Cryo-negatively stainedimages of two-dimensional VacAcrystalline arrays. Left panels: thearrays appear to consist of repeat-ing patterns of large hexagonalrings interspersed regularly withsmaller rings. Right panel: roughimage-averaging of the crystals onthe left, done by repeated image-dislocation and superimpositionusing Adobew Photoshop. Thelarge hexagonal rings and smallerjunctional rings are linked byfaintly visible pairs of connectors.The smaller rings are formed byinterdigitation of six of theseconnectors, emanating from threeadjacent crystalline units. Magnifi-cation, 430,000 £ .


spikes rather than as petals.18,19 At first, one mightsuspect that this difference was an artifact due toexposure to ammonium molybdate. However, inone instance, we imaged VacA by cryo-EM withthe molecules suspended in pure water ratherthan in our usual negative stain, and the peripheralcomponents of VacA still appeared as narrowspikes. Furthermore, in another instance, weimaged VacA by traditional negative staining (air-drying in uranyl acetate), and again the peripheralcomponents appeared as narrow spikes. Hence,the basis for the differences in VacA appearanceresulting from these two different EM approachesremains somewhat unclear. The most straight-forward explanation is that they simply reflect thefundamental differences between a surface-imageof a molecule under a relatively electron-opaqueplatinum replica, versus a relatively “holistic”image of a molecule suspended in a semi-translucent film of negative stain.

CryoEM imaging of 2D VacA crystal formation

In addition to isolated VacA oligomers, cryo-negatively stained preparations occasionally dis-played clusters of closely packed VacA oligomersthat appeared to interdigitate with each other viacomplex lateral associations (Figure 7). Thesesmall 2D crystalline arrays consisted of hexagonalrings interspersed regularly with smaller rings, allin a triagonal array. Analysis of Photoshop-averaged images of such crystals (Figure 7, rightpanel) suggested that the smaller, ring-likedomains could represent an overlap of the petalsor spikes from three adjacent VacA dodecamers.One fundamental difference between these twoforms, however, was that the peripheral spikes ofthe isolated dodecamers generally appeared assingular rods, whereas they appeared as pairedstructures in the 2D crystals (cf. Figure 7, especiallythe reconstruction on the right.) However, uponcloser inspection of our complete set of images ofindividual VacA oligomers, a few were found inwhich one or more of the peripheral spikesappeared to be composed of pairs of rodlets (cf.Figure 4, row 4, fourth image). With this in mind,we reviewed all our images and found a few more

oligomers in which all of the peripheral spikesappeared to be composed of such “paired” struc-tures, resulting in forms that exhibited 12 visibleperipheral rodlets in a “Star of David” configur-ation (Figure 8). Possibly, this partial dissociationof the peripheral spikes in otherwise intactdodecamers reflected a general tendency of VacAto dissociate during cryo-negative staining. How-ever, a similar form of partial dissociation wasobserved occasionally by deep-etch EM (data notshown). In any case, these partially dissociated“intermediate forms” match the appearance of the“unit cell” of the 2D VacA crystals so closely thatthey further support the conclusion that the crys-tals are “two layers thick”, i.e. that they are indeedcomposed of interdigitated dodecamers. Thesecrystals look almost exactly like the 2D crystals ofVacA seen previously by AFM;9 however, inthat study they were interpreted as being“single-layered”. This discrepancy is addressed inDiscussion.

Deep-etch EM analysis of VacA bound tomembrane surfaces

Ultimately, the most relevant oligomeric form ofVacA should be the one that interacts with bio-logical membranes. As a first approach towarddetermining what this form might be, weattempted to image the interaction of VacA witheukaryotic cells by deep-etch EM. Unfortunately,we found it technically unfeasible to image VacAon the surfaces of several different types of epi-thelial cells, simply because these cells displayedsuch “bumpy” surfaces to start with. Conse-quently, we chose to use rabbit erythrocytes as amodel system, because we knew from previouswork that the surfaces of these cells appearsexceedingly smooth by deep-etch EM. Soon, wefound that we needed to use acid-activated VacAto obtain a significant amount of binding of VacAto such erythrocytes. Thereupon, deep-etch EMrevealed that acid-activated VacA bound to thesurface of rabbit erythrocytes appeared as veryfaint, very shallowly raised flower-shaped struc-tures, comprised of central rings surrounded bysix or seven relatively globular petals (Figure 9,

Figure 8. Cryo-negatively stainedimages of VacA oligomers with avariant morphology. Top row:examples of symmetrical complexesthat display the usual six arms, likethe complexes shown in Figure 4.Bottom row: symmetrical com-plexes that display up to 12 arms,possibly reflecting some degree ofVacA disassembly. (In the thirdpanel, an image of the “Star ofDavid” is superimposed upon theright-most panel, for illustrativepurposes.) Magnification, 300,000 £ .


top two rows). These structures closely resembledthe counterclockwise thin-forms of VacA seen onlyvery occasionally on mica (Figure 1, row 2).

Besides binding VacA to natural membranes, weattempted to characterize the interaction of VacAwith other sorts of hydrophobic surfaces, by bind-ing it to artificial liposomes. To do this, we firstadsorbed liposomes to mica and then incubatedthe adsorbed liposomes with acidified VacA at pH3.5. Following neutralization of the pH and wash-ing, the samples were prepared for deep-etch EM.Deep-etch EM images of these samples (Figure 9,bottom two rows) again revealed relatively flatforms of VacA oligomers, and these again consist-ently exhibited a counterclockwise chirality and aprominent central ring. In both cases, then, weinterpret these oligomeric structures to representthe thin, single-layered hexameric form of VacA,bound to either erythrocyte membranes or to lipo-somes by its hydrophobic surface.

Discussion

The images of cryo-negatively stained VacA pre-sented here help to resolve several differences ininterpretation that have arisen in previous studiesthat utilized solely deep-etch EM or AFM.9,18 – 20

Foremost, they confirm that VacA has the capacityto assemble into several different types of oligo-meric structure, including: (i) well-contrasted,symmetrical flowers that appear as bilayered struc-tures when viewed on edge; (ii) less well-contrasted astral arrays that display distinctchirality; and (iii) small 2D crystals composed ofinterdigitated oligomers apparently constructedlike form (i) above. The first two of these formsmatch closely the VacA oligomers recognizedearlier by deep-etch EM.18,19 That is, the well-contrasted symmetrical arrays seen by cryo-negative staining (Figure 4) correspond to thesymmetrical flowers seen abundantly by deep-etch EM (Figure 1, row 1). Likewise, the less well-contrasted, chiral arrays seen by cryo-negativestaining (Figure 5) correspond to the thin-formsseen previously by deep-etch EM (Figure 1, row3). In the model we proposed previously,19 theVacA flower was interpreted as being a sandwichof two such thin-forms adhering to each other viatheir hydrophobic surfaces (Figure 2). Striking sup-port for this interpretation comes now from theside-views of VacA oligomers seen in the presentstudy (Figure 6), wherein they do indeed look dis-tinctly bilayered. The tightest apposition in thesesandwiches appears to occur between the centralring-like domains (Figure 6). Hence, we imagine

Figure 9. 3D anaglyph stereo view of VacA bound to rabbit erythrocytes or liposomes and imaged by deep-etch EM.Top two rows: VacA was acid-activated and then reneutralized before incubation with erythrocytes for five minutes.After washing at neutral pH, bound VacA molecules were imaged by quick-freezing the erythrocytes, followed byfreeze-fracturing and deep-etching them. Bottom two rows: VacA was acidified to pH 3.5 by addition of glycine buffer,and then incubated for two minutes with liposomes that had been pre-adsorbed to a mica substrate. Most of the com-plexes bound to the erythrocytes or liposomes appeared relatively “flat”, yet still displayed a distinct central ring andexhibited a counterclockwise chirality, similar to the chiral flat forms seen in Figure 1, row 2. Magnification, 300,000 £ .


that these regions must be the most hydrophobicportions of the complex. In contrast, the petals insuch side-views typically appear to be slightlyseparated from each other, suggesting that theseregions may be relatively hydrophilic. The factthat cryo-negative staining promotes a ratherdramatic overall dissociation of VacA is furtherdiscussed below, but in terms of determining thebasic construction of VacA, it is relevant to mentionhere that cryo-negative staining yields occasionalen face views of oligomers in which the peripheralpetals have clearly split into pairs (Figure 8). More-over, pairs of connections are displayed in eachunit-cell of the 2D crystal (Figure 7). Both of theseobservations are critical pieces of evidence infavor of the view that the typical VacA flowersuspended in an aqueous environment is a sand-wiched dodecamer (Figure 2).

Nevertheless, despite the fact that the currentcryo-negatively stained images provide criticalsupport for our previous interpretation of VacAconstruction,19 they display some significant differ-ences from earlier deep-etch EM images. Onedifference is in the appearance of the peripheralparts of the thicker oligomers. By cryo-negativestaining, these look more spike-like than thebulbous or petal-like images of deep-etch EM. Wepresume that their true structure must lie some-where in between these two extremes. The plati-num replication required for deep-etch EM mustexaggerate the breadth of these structures, whereasthe negative staining probably fills them in (orpositively stains their low-density regions) tosome extent. Furthermore, the overall translucencyof negative-stain images, in general, undoubtedlycontributes to their relatively delicate, spike-likeappearance. Additionally, it is quite likely that thepetals have greater mobility in free solution thanwhen adsorbed to a solid surface like mica, andhence are free during negative staining to undergovarious translations and rotations. Still, despitetheir relatively “spike-like“ appearance, the distalappendages in cryo-EM images often do appear tohave slightly higher-contrast regions at their tips(Figure 5). (This is clearly apparent in recent opti-cal reconstructions of cryo-negatively stainedVacA that are currently being prepared for sub-sequent publication.) These widened distal regionsmay be free to rotate during negative staining, suchthat they are rarely visible in the direction of sight.Indeed, such mobility at a “hinge” near the baseof the petals could be important in forming thesandwiched bilayer, as well as in forming thebroader 2D crystalline arrays.

Other discrepancies between the two differentpreparative approaches include the fact that thecurrent cryo-EM images show a much higher pro-portion of chiral thin-forms than of symmetricalflowers (Figure 3). Also, the cryo-EM images showa relatively high concentration of putative VacAmonomers in the background (Figure 3). Thesedifferences do not appear to be due to systematicchanges in sample preparation over time, since

the very same preparation of VacA showed thesedifferences when prepared simultaneously byboth techniques. Hence, there seems to be some-thing about the cryo-negative stain technique thattends to dissociate VacA. Possibly, it may just bethe high cation concentration in the negative stain(15–30% NH4

þ), since recent work has shown thatVacA is dissociated by either high or low pHvalues.19,22,23 In this regard, it is interesting to notethat the air/water interface present during cryo-negative staining tends to promote a re-associationof VacA monomers into the thinner, chiral forms.Likewise, other hydrophobic surfaces like those ofthe liposomes and erythrocytes used here appearto promote a re-association of acid-dissociatedVacA into the thinner, chiral forms. This may bemirroring the mechanism by which VacA generatesion-selective pores or channels in planar lipidbilayers and eukaryotic cell membranes.9 – 12

Another difference between the two techniquesbeing compared here is that the handedness of thechiral thin-forms is generally opposite in therespective images: clockwise by deep-etch EM andcounterclockwise by cryo-EM. Earlier, we inter-preted the clockwise orientation seen by deep-etchEM as resulting from a preferential adherence ofVacA to the mica, due to one of its surfaces havinga higher affinity for mica than the other.19 Specifi-cally, we argued that this would be their hydro-philic surface, simply because the mica itself ishighly charged. Now, if they adhered to mica bytheir hydrophilic surfaces (as, in fact we imaginewhole dodecamers do, as well) then this wouldtend to expose for replication and viewing therelatively hydrophobic surfaces of the chiralforms. According to our model, above, this wouldbe the surface located inside the VacA sandwichwhenever two such thin-forms merged to form athicker, flower-like dodecamer.

The present cryo-images of chiral thin-formslook almost identical with their deep-etch counter-parts, except that they display this oppositehandedness. At first sight, this might seem to posea problem for our model of VacA construction;but, in fact, it strongly substantiates the model.Our reasoning here is as follows: the chiral thinforms seen by cryo-EM can be recognized byvarious clues, including stereo-EM, as being VacAoligomers that have adhered to the air–water inter-face that is present invariably during cryo-negativestaining. Now, the relatively more hydrophobicsurface of VacA would be expected to adhere toan air/water interface, simply because such a sur-face is much more hydrophobic than mica; and ithappens that in our laboratory, the air–water inter-face in cryo-EM faces the bottom of the EM grid(and hence, faces down in the electron micro-scope). (This is simply because this untouchedair–water interface faces away from the surfacethat is blotted in the moments just before plunge-freezing.27) By contrast, in deep-etch EM, moleculesadsorbed to the mica always face upward in themicroscope, simply because deep-etch replicas are


picked up onto grids from below, after they havebeen cleaned. Hence, the only reason that theintensely chiral “flat forms” seen by cryo-EM lookopposite in handedness to their deep-etch EMcounterparts, is simply because they weremounted the other way up in the electronmicroscope.

We are left to ponder why the two surfaces ofhemi-VacA should display such distinct differencesin their affinity for air versus mica. This will prob-ably be answered only by future advances in thegeneral understanding of the physical chemistryof surface-adsorption of macromolecules. How-ever, in any case, we may properly conclude thatthere are several reasons for believing that thechiral VacA thin-form is a polar structure, withone side relatively hydrophobic and the other siderelatively hydrophilic, and that this polaritydictates VacA’s pattern of self-assembly as well asits propensity to adsorb to various differentsubstrates.

The repeating units of the 2D VacA crystalsvisualized in this study each contain a prominenthexagonal central ring, and in this respect appearto be built out of individual VacA dodecamers(compare Figures 1, 4, and 7). However, they differfrom isolated VacA dodecamers in two importantrespects: (i) in the 2D crystals, six small peripheralrings are visible surrounding each central ring,whereas no such peripheral rings are visible in theisolated dodecamers; and (ii) each repeating unitof the crystal contains 12 visible arms or connec-tors, whereas only six petals or spikes are visiblein most of the isolated dodecamers. We proposethat the presence of 12 visible arms in the 2D crys-tals results from a “splaying” of VacA petals, asshown in Figures 8 and 10(a). Overlap or inter-digitation of the distal parts of the arms fromadjacent dodecamers would give rise to the smallrings that appear amidst the larger hexagonalrings (Figure 10(b) and (c)). (Specifically, thiswould be an interdigitation of six arms arisingfrom three adjacent crystalline units.) One mightreasonably presume that such interdigitationwould be associated with extensive domain-

swapping and major changes in the conformationof the component VacA monomers.28

The overall capacity of VacA to assemble into avariety of oligomeric structures is presumablydirectly relevant to the process by which this toxininserts into lipid membranes to form anion-conductive channels.9 – 13 Acid-treatment of VacAresults in disassembly of dodecamers, and suchdisassembly is associated with a marked enhance-ment in the capacity of VacA to bind to mem-branes, to form membrane channels, and to exertcytotoxic effects on eukaryotic cells.9,19,21 –24. On thebasis of these observations, it seems likely thatVacA dodecamers are relatively inactive comparedto the dissociated forms of VacA.

The question thus becomes: what is the activepore-forming configuration that VacA assumeswhen bound to natural membranes? Prior to thisstudy, VacA bound to membranes had beenimaged only by AFM, and only after it hadreannealed into relatively large hexamericcrystals.9 Individual oligomers that might reason-ably represent individual pores were not success-fully imaged by AFM.9 On the basis of the veryshallow elevation of the VacA crystals observedby AFM, it was proposed that VacA had assembledin that case in a singled-layered form.9 However,the 2D crystals seen in the present study lookexactly the same as those seen by AFM, but areclearly bilayers of VacA. (This we can concludewith some certainty, since their electron-density isjust as great as the bilayered, dodecameric formsof VacA seen by cryo-EM.) This raises the concernthat the crystals observed previously by AFM mayhave been partially flattened (and/or partiallyembedded in lipid) in the course of that study. Inany case, in the current study we were able toobtain deep-etch EM images of individual com-plexes of acid-dissociated and reannealed VacAbound to the surface of erythrocytes and lipo-somes, images more likely to approximate theappearance of the natural VacA “pore” in situ.These individual oligomeric complexes lookedvery much like the counterclockwise chiral formsseen very rarely on mica (Figure 1, row 2). We

Figure 10. Structural model for the growth of 2D VacA crystals via interdigitation of individual dodecamers. Thiswould require a “splaying” of the normally overlapped arms in each dodecamer, followed by a reannealing of the dis-tal portions of six arms from three adjacent dodecamers to form each of the small rings found in-between the largerrings (cf. Figure 7).


presume that these forms resulted from thereannealing of monomeric VacA into oligomersafter adsorption onto the hydrophobic lipid sur-face, similar to the reannealing seen in cryo-nega-tively stained preparations at the air/waterinterface. The capacity of VacA to form a varietyof chiral thin forms, ranging from six to nine-membered rings, suggests that VacA might be ableto form a variety of pore-sizes when acting withinnatural membranes. This will be important toinvestigate in future work.

Our interpretation of how VacA interacts withmembranes shares several features in commonwith the current views of how the bacterial toxinaerolysin forms membrane channels. Aerolysinexists in solution primarily in the form of dimers,but forms heptameric ring-shaped complexeswhen attached to biological membranes.29,30 Thus,with both VacA and aerolysin, the process of poreformation seems to involve disassembly of water-soluble complexes, binding of monomeric toxin tothe membrane and, finally, reassembly of oligo-meric structures on the membrane surface. Furtherstudies of VacA structure, oligomerization, andchannel formation are certain to provide importantinsights into the mechanisms by which this toxinalters the structure and function of eukaryotic cells.

Materials and Methods

Preparation of H. pylori VacA

H. pylori strain 60190 (American Type Culture Collec-tion 49503, Rockville, MD), was cultured in ambient aircontaining 6% (v/v) CO2 for 48 hours at 37 8C in sulfite-free Brucella broth containing 0.5% (w/v) charcoal.After centrifugation of the culture at 10,000g for15 minutes, supernatant proteins were precipitated witha 50% saturated solution of ammonium sulfate andresuspended in phosphate-buffered saline.14 VacA waspurified by a two-step approach, involving affinityadsorption to Matrex cellufine sulfate beads (Millipore,Bedford, MA), followed by gel-filtration chromatographyusing a Superose 6 16/50 column (Pharmacia Biotech,Piscataway, NJ).14,19,31 Alternatively, in some experimentsVacA was purified by gel-filtration chromatographyalone.

Cryo-negative staining

Purified VacA oligomers were imaged by cryo-negative staining as described.26,27,32 Briefly, a drop ofprotein solution was placed onto a grid supporting agold-coated, holey, carbon film. After 30 seconds, thegrid was floated for one minute on a 16% (w/v)ammonium molybdate solution at neutral pH. This lefta suspension of protein oligomers within the holes ofthe carbon film but replaced the medium by themolybdate stain. The drop on the grid was then blottedrapidly to near-dryness, leaving only a thin film of liquidwithin the holes, and the grid was allowed to fall into acryogenic liquid such as liquid propane or ethane,where it vitrified within about 1026 second. Keeping thetemperature at about 2180 8C, the specimen was thentransferred into a Gatan 626 cryo-specimen holder and

mounted in a cryo-electron microscope (Philips CM 12cryo). Observation was made at low temperature withminimum electron dose. Images were recorded onphotographic film or with a digital CCD camera (Gatanmodel 794). Those photographed on film were recordedinitially at a magnification of 44,000 £ on Kodak S0163film and then transferred to a computer via a digitalcamera mounted on a standard copy stand, to give afinal magnification of 300,000 £ .33 Then, using AdobePhotoshop, individual molecules were chosen for photo-montages, and averaged views were obtained usingstandard Photoshop layering software. Finally, stereoanaglyphs were prepared as described.34

Deep-etch electron microscopy of VacA bound tomembrane surfaces

To decorate artificial lipid surfaces with VacA, lipo-somes were first prepared from a chloroform mixture of30% (w/v) 1,2-dioleoyl-sn-glycero-3-phosphate and 70%(w/v) 1,2,-dioleoyl-sn-glycero-3-phosphocholine (AvantiPolar Lipids). After evaporation of chloroform under N2

gas, this mixture was hydrated with an aqueous solutionof 10 mM Hepes (pH 7.4), 10 mM KCl, 1 mM EDTA. Thissolution was freeze-thawed five times and then soni-cated for five 30 second pulses in a bath sonicator(1600 W). The resultant lipid vesicles were sized bypassage through 200 nM polycarbonate filters (NuclearPore, Pleasanton, CA). Liposome size was determinedby dynamic light-scattering using a N4MD submicronparticle analyzer (Colter, Hialeaha, FL). Liposomes wereapplied to mica flakes at 100 mg/ml and allowed toadsorb for one minute, after which the mica was washedin the glycine buffer used to acidify and dissociate VacA,described next. VacA was acidified to pH 3.5 by adding200 mM glycine buffer (pH 3.5) containing 70 mM KCland 2 mM MgCl2, and then was added to the mica-attached liposomes in the same buffer and incubated fortwo minutes. The specimens were then neutralized bywashing the mica in 70 mM KCl, 30 mM Hepes (pH7.2), 2 mM MgCl2. After a further 15 minutes ofincubation at room temperature, the specimens werefinally quick-frozen in preparation for deep-etch EM,which was done exactly as described.19

To decorate erythrocytes with VacA, fresh rabbit bloodwas used. The erythrocytes were washed free of serumwith a standard PBS and attached to polylysine-treatedglass coverslips, as described.35 After washing awaynon-adherent cells, the coverslips were exposed to VacAunder various conditions. When oligomeric VacA wasapplied at neutral pH, little or no binding could beobserved. When acid-disassembled forms of VacA wereapplied at pH 3.5, complete saturation of the erythrocytesurface with VacA seemed to occur, but the erythrocyteswere so deformed by the low pH that the resultant EMimages were almost uninterpretable. Hence, a “middleroad” was chosen, in which we preacidified a solutionof VacA oligomers with dilute HCl, left it at this pH for30 minutes at 37 8C, and then diluted it into a neutralbuffer immediately before presenting it to the erythro-cytes. From previous work, we know that such reneutra-lization under dilute conditions leads to only a very slowre-annealing of VacA into oligomers.19 Hence, we pre-sume this VacA was largely still monomeric whenapplied to the erythrocytes. In all cases, exposure to theerythrocytes was for five minutes in PBS, followed bytwo quick PBS rinses and prompt quick-freezing of the


erythrocyte-laden coverslips in preparation for deep-etchEM, as above.19

Acknowledgments

This work was supported by grants from the NIH toJ.H. (GM #29647) and to T.C. (AI39657 and DK53623),and by the Medical Research Service of the Departmentof Veterans Affairs. We thank Robyn Roth for producingall the deep-etch replicas used in this study, JenniferScott and Hiroshi Morisaki for the computer workinvolved in generating the Figures, and Beverly Hossefor assistance with VacA purification.

References

1. Cover, T. L., Berg, D. E., Blaser, M. J. & Mobley, H. L.T. (2001). Helicobacter pylori pathogenesis. InPrinciples of Bacterial Pathogenesis (Groisman, E A.,ed.), pp. 509–558, Academic Press, San Diego, CA.

2. Atherton, J. C., Cover, T. L., Papini, E. & Telford, J. L.(2001). Vacuolating cytotoxin. In Helicobacter pylori:Physiology and Genetics (Mobley, H L T., Mendz, G L.& Hazell, S L., eds), pp. 97–110, ASM Press,Washington, DC.

3. Montecucco, C., Papini, E., de Bernard, M., Telford,J. L. & Rappuoli, R. (1999). Helicobacter pylori vacuo-lating cytotoxin and associated pathogenic factors.In The Comprehensive Sourcebook of Bacterial ProteinToxins (Alouf, J E. & Freer, J H., eds), pp. 264–286,Academic Press, San Diego, CA.

4. Molinari, M., Galli, C., Norais, N., Telford, J. L.,Rappuoli, R., Luzio, J. P. & Montecucco, C. (1997).Vacuoles induced by Helicobacter pylori toxin containboth late endosomal and lysosomal markers. J. Biol.Chem. 272, 25339–25344.

5. Satin, B., Norais, N., Telford, J., Rappuoli, R., Murgia,M., Montecucco, C. & Papini, E. (1997). Effect ofHelicobacter pylori vacuolating toxin on maturationand extracellular release of procathepsin D and onepidermal growth factor degradation. J. Biol. Chem.272, 25022–25028.

6. Molinari, M., Salio, M., Galli, C., Norais, N.,Rappuoli, R., Lanzavecchia, A. & Montecucco, C.(1998). Selective inhibition of Ii-dependent antigenpresentation by Helicobacter pylori toxin VacA. J. Exp.Med. 187, 135–140.

7. Papini, E., Satin, B., Norais, N., de Bernard, M.,Telford, J. L., Rappuoli, R. & Montecucco, C. (1998).Selective increase of the permeability of polarizedepithelial cell monolayers by Helicobacter pylorivacuolating toxin. J. Clin. Invest. 102, 813–820.

8. Papini, E., Satin, B., Bucci, C., deBernard, M., Telford,J. L., Manetti, R., Rappuoli, R., Zerial, M. &Montecucco, C. (1997). The small GTP bindingprotein rab7 is essential for cellular vacuolationinduced by Helicobacter pylori cytotoxin. EMBO J. 16,15–24.

9. Czajkowsky, D. M., Iwamoto, H., Cover, T. L. & Shao,Z. (1999). The vacuolating toxin from Helicobacterpylori forms hexameric pores in lipid bilayers at lowpH. Proc. Natl Acad. Sci. USA, 96, 2001–2006.

10. Iwamoto, H., Czajkowsky, D. M., Cover, T. L., Szabo,G. & Shao, Z. (1999). VacA from Helicobacter pylori: ahexameric chloride channel. FEBS Letters, 450,101–104.

11. Szabo, I., Brutsche, S., Tombola, F., Moschioni, M.,Satin, B., Telford, J. L. et al. (1999). Formation ofanion-selective channels in the cell plasma mem-brane by the toxin VacA of Helicobacter pylori isrequired for its biological activity. EMBO J. 18,5517–5527.

12. Tombala, F., Carlesso, C., Szabo, I., de Bernard, M.,Reyrat, J. M., Telford, J. L. et al. (1999). Helicobacterpylori vacuolating toxin forms anion-selectivechannels in planar lipid bilayers: possible impli-cations for the mechanism of cellular vacuolation.Biophys. J. 76, 1401–1409.

13. Vinion-Dubiel, A., McClain, M. S., Czajkowsky, D. M.,Iwamoto, H., Ye, D., Cao, P. et al. (1999). A Dominantnegative mutant of Helicobacter pylori vacuolatingtoxin (VacA) inhibits VacA-induced cell vacuolation.J. Biol. Chem. 274, 37736–37742.

14. Cover, T. L. & Blaser, M. J. (1992). Purification andcharacterization of the vacuolating toxin fromHelicobacter pylori. J. Biol. Chem. 267, 10570–10575.

15. Cover, T. L., Tummuru, M. K. R., Cao, P., Thompson,S. A. & Blaser, M. J. (1994). Divergence of geneticsequences for the vacuolating cytotoxin amongHelicobacter pylori strains. J. Biol. Chem. 269,10566–10573.

16. Schmitt, W. & Haas, R. (1994). Genetic analysis of theHelicobacter pylori vacuolating cytotoxin: structuralsimilarities with the IgA protease type of exportedprotein. Mol. Microbiol. 12, 307–319.

17. Telford, J. L., Ghiara, P., Dell’Orco, M., Comanducci,M., Burroni, D., Bugnoli, M. et al. (1994). Gene struc-ture of the Helicobacter pylori cytotoxin and evidenceof its key role in gastric disease. J. Exp. Med. 179,1653–1658.

18. Lupetti, P., Heuser, J. E., Manetti, R., Massari, P.,Lanzavecchia, S., Bellon, P. L. et al. (1996). Oligomericand subunit structure of the Helicobacter pylori vacuo-lating cytotoxin. J. Cell. Biol. 133, 801–807.

19. Cover, T. L., Hanson, P. I. & Heuser, J. E. (1997). Acid-induced dissociation of VacA, the Helicobacter pylorivacuolating cytotoxin, reveals its pattern of assembly.J. Cell. Biol. 138, 759–769.

20. Lanzavecchia, S., Bellon, P. L., Lupetti, P., Dallai, R.,Rappuoli, R. & Telford, J. (1998). Three-dimensionalreconstruction of metal replicas of the Helicobacterpylori vacuolating cytotoxin. J. Struct. Biol. 121, 9–18.

21. deBernard, M., Papini, E., deFilippis, V., Gottardi, E.,Telford, J., Manetti, R. et al. (1995). Low pH activatesthe vacuolating toxin of Helicobacter pylori, whichbecomes acid and pepsin resistant. J. Biol. Chem. 270,23937–23940.

22. Molinari, M., Galli, C., de Bernard, M., Norais, N.,Ruysschaert, J-M., Rappuoli, R. & Montecucco, C.(1998). The acid activation of Helicobacter pylori toxinVacA; structural and membrane binding studies.Biochem. Biophys. Res. Commun. 248, 334–340.

23. Yahiro, K., Niidome, T., Kimura, M., Hatakeyama, T.,Aoyagi, H., Kurazono, H. et al. (1999). Activation ofHelicobacter pylori VacA toxin by alkaline or acid con-ditions increases its binding to a 250-kDa receptorprotein-tyrosine phosphatase beta. J. Biol. Chem. 274,36693–36699.

24. McClain, M. S., Schraw, W., Ricci, V., Boquet, P. &Cover, T. L. (2000). Acid activation of Helicobacterpylori vacuolating toxin (VacA) results in toxininternalization by eukaryotic cells. Mol. Microbiol.37, 433–442.

25. Reyrat, J. M., Lanzavecchia, S., Lupetti, P., deBernard, M., Pagliaccia, C., Pelicic, V. et al. (1999).


3D imaging of the 58 kDa cell binding subunit of theHelicobacter pylori cytotoxin. J. Mol. Biol. 290,459–470.

26. Adrian, M., Dubochet, J., Fuller, S. D. & Harris, J. R.(1998). Cryo-negative staining. Micron, 29, 145–160.

27. Dubochet, J., Adrian, M., Lepault, J. & McDowell,A. W. (1985). Cryo-electron microscopy of vitrifiedbiological specimens. Trends Biochem. Sci. 10,143–146.

28. Schlunegger, M. P., Bennett, M. J. & Eisenberg, D.(1997). Oligomer formation by 3D domain swapping:a model for protein assembly and misassembly.Advan. Protein Chem. 50, 61–122.

29. Wilmsen, H. U., Leonard, K. R., Tichelaar, W.,Buckley, J. T. & Pattus, F. (1992). The aerolysin mem-brane channel is formed by heptamerization of themonomer. EMBO J. 11, 2457–2463.

30. Fivaz, M., Velluz, M.-C. & van der Goot, F. G. (1999).Dimer dissociation of the pore-forming toxin

aerolysin precedes receptor binding. J. Biol. Chem.274, 37705–37708.

31. Manetti, R., Massari, P., Burroni, D., de Bernard, M.,Marchini, A., Olivieri, R. et al. (1995). Helicobacterpylori cytotoxin: importance of native conformationfor induction of neutralizing antibodies. Infect.Immun. 63, 4476–4480.

32. Dubochet, J., Adrian, M., Chang, J. J., Homo, J. C.,Lepault, J., McDowell, A. W. et al. (1988). Cryo-electron microscopy of vitrified specimens. Q. Rev.Biophys. 21, 129–228.

33. Heuser, J. E. (2000). How to convert a traditionalelectron microscopy laboratory to digital imaging:follow the ‘middle road’. Traffic, 1, 614–621.

34. Heuser, J. E. (2000). Membrane traffic in anaglyphstereo. Traffic, 1, 35–37.

35. Heuser, J. E. (2000). The production of ‘cell cortices’for light and electron microscopy. Traffic, 1, 545–552.

Edited by A. Klug

(Received 24 August 2001; received in revised form 9 January 2002; accepted 7 February 2002)


multiple oligomeric states of the helicobacter pylori vacuolating toxin demonstrated by...

Documents