1739

Journal of Cell Science 109, 1739-1747 (1996)Printed in Great Britain © The Company of Biologists Limited 1996JCS4226

Vaccinia virus: a model system for actin-membrane interactions

Sally Cudmore, Inge Reckmann, Gareth Griffiths and Michael Way*

Cell Biology Programme, EMBL, Meyerhofstrasse 1, Heidelberg 69117, Germany

*Author for correspondence (e-mail: way@embl-heidelberg.de)

Our understanding of the interactions between the actincytoskeleton and cellular membranes at the molecular levelis rudimentary. One system that offers an opportunity toexamine these interactions in greater detail is provided byvaccinia virus, which induces the nucleation of actin tailsfrom the outer membrane surrounding the virion. Tofurther understand the mechanism of their formation andhow they generate motility, we have examined the structureof these actin tails in detail. Actin filaments in vaccinia tailsare organized so they splay out at up to 45° from the centreof the tail and are up to 0.74 µm in length, which is con-siderably longer than those reported in the Listeria system.Actin filaments show unidirectional polarity with theirbarbed filament ends pointing towards the surface of thevirus particle. Rhodamine-actin incorporation experiments

show that the first stage of tail assembly involves apolarized recruitment of G-actin, and not pre-formed actinfilaments, to the membrane surrounding the virion. Incor-poration of actin into the tail only occurs by nucleationfrom the viral surface, suggesting filament ends in the tailare blocked against further actin addition. As virusparticles fuse with the plasma membrane during theextention of projections, actin nucleation sites previously inthe viral membrane become localized to the plasmamembrane, where they are able to nucleate actin polymer-ization in a manner analogous to the leading edge of motilecells.

Key words: Vaccinia virus, Actin tail, Membrane

SUMMARY

INTRODUCTION

A key mechanism in the virulence of any cellular pathogen isits ability to move from one cell to another to facilitate thespread of infection. A group of unrelated bacterial pathogens,Shigella, Listeria and Rickettsia, that are the causative agentsof a number of important diseases including meningitis, septi-caemia, bacillary dysentery and spotted fever, have indepen-dently developed a similar actin-based mechanism that isessential for their intercellular spread (for reviews see Tilneyand Tilney, 1993; Cossart and Kocks, 1994; Cossart, 1995).These bacteria are not the only infectious agents that haveprofound effects on the actin cytoskeleton of their host. Severalviruses have been shown to disrupt or stabilize the actincytoskeleton (Tyrell and Norrby, 1978; Giuffre et al., 1982;Murti et al., 1985; Bohn et al., 1986). The best studied exampleis that of vaccinia virus, the prototype member of the orthopoxgenus and a close relative of smallpox (Stokes, 1976; Hiller etal., 1979, 1981; Krempien et al., 1981; Cudmore et al., 1995).

Vaccinia, one of the largest and most complex virusesknown, has a programmed series of interactions with the hostcell that involve wrapping by two membrane cisternae ofdifferent subcellular origin during viral assembly and matura-tion. An intriguing characteristic of the vaccinia virus assemblyprocess is that it results in two different infectious forms. Thefirst of these, the intracellular mature virus (IMV), is sur-rounded by a membrane cisterna derived from the intermedi-ate compartment (Sodeik et al., 1993). Infectious IMV are

released when the cell lyses due to the cytotoxic effects ofinfection. Alternatively, a small proportion of IMV, whichvaries according to virus strain and cell type, can undergo asecond wrapping step by a cisternal domain derived from thetrans-Golgi network (Payne, 1980; Payne and Kristensson,1985; Schmelz et al., 1994). This form is called the intracellu-lar enveloped virus (IEV) and escapes from the cell by fusionof its outer membrane with the plasma membrane of the host,thereby releasing the second infectious form called the extra-cellular enveloped virus (EEV) (Dales, 1971; Morgan, 1976;Payne, 1980; Blasco and Moss, 1991). During the fusion ofIEV with the plasma membrane, a small proportion of EEVparticles are not released into the medium but remain associ-ated with the outside of the cell. These particles are refered toas cell-associated enveloped virus (CEV) (Blasco and Moss,1992).

The first indication that vaccinia virus was able to interactwith the cytoskeleton during its complex assembly processcame from high voltage electron microscopy studies whichshowed virus particles at the tips of large microvilli-like pro-jections in infected cells (Stokes, 1976). Subsequent studiesconfirmed that these vaccinia-tipped projections containedactin, as well as the actin cross-linking proteins α-actinin,filamin and fimbrin but not tropomyosin or myosin (Hiller etal., 1979, 1981). Furthermore, inhibition of virus assemblyprevented the formation of the large microvilli suggesting thatthese structures were virally induced (Krempien et al., 1981).

In light of the effects of bacterial pathogens on the actin

1740 S. Cudmore and others

cytoskeleton we have recently re-examined the effects ofvaccinia virus on the actin cytoskeleton (Cudmore et al., 1995).Infection by vaccinia virus not only results in the formation ofvirally-induced microvilli but also in the disruption of actinstress fibres and the formation of actin tails in the cytoplasmof the host cell that are very reminiscent of those seen inbacterial systems. Studies using a mutant virus which cannotform IEV, as well as infection with wild-type virus in theprescence of a drug which prevents IEV formation, demon-strated that it is the IEV that recruits actin to form tails.Vaccinia virus uses these actin tails to move around in thecytoplasm of the host cell at a speed of 2.8 µm/min and toproject into and infect neighbouring cells (Cudmore et al.,1995). The existence of actin-based motility of vaccinia virussuggests that diverse cellular pathogens have developed acommon mechanism to exploit the host actin cytoskeleton asa means to facilitate their spread between cells.

In order to learn more about the process of actin tailmediated motility, we have analyzed the structure of actinfilaments in tails induced by vaccinia virus. Using rhodamineactin incorporation into permeabilized cells, we have shownthat nucleation and growth of actin filaments occurs in apolarized fashion from the viral surface. Furthermore, theappearance of the actin filaments in the tails nucleated by theviral membrane is strikingly similar to the organization andpolarity of actin filaments at the leading edge of motile cells.

MATERIALS AND METHODS

Immunofluorescence microscopyHeLa cells were grown, infected and fixed at 8 hours post infection(p.i.) as previously described (Cudmore et al., 1995). Infected cellswere labelled with anti-14kD viral antibodies (Rodriguez et al., 1985)and either rhodamine or bodipy phalloidin (Molecular Probes,Eugene, OR, USA). For maximium resolution, all immunofluoresenceexperiments were examined and recorded with ×63 or ×100 lensesusing a Zeiss Axiophot microscope. The images in Figs 1 and 2 weregenerated by scanning prints generated from negatives on a flat bed-scanner to convert images into a digital format. These digital imageswere enlarged and annotated using the Adobe 3.0 software package.

Electron microscopyHeLa cells were grown and infected with vaccinia virus strain WR ata multiplicity of infection of 1 pfu/cell as described (Doms et al.,1990). For S1 myosin labelling, infected cells were washed at 16hours p.i. with ice cold PBS and incubated with 5 mg/ml S1 myosinin 0.1 M phosphate buffer (PB), pH 6.8, and 0.2 mg/ml saponin for30 minutes on ice. The cells were washed twice with PB and subse-quently fixed in 1% gluteraldehyde, 2% tannic acid and 50 mMphosphate buffer, pH 6.8, for 30 minutes at room temperature. Afteranother wash in PB, the cells were post fixed in 1% OsO4 and 1.5%KFe(Cn)6 for 30 minutes on ice in the dark. Following three rinses indistilled water, the samples were dehydrated in 50% ethanol andstained overnight in 70% ethanol saturated with uranyl acetate. Thefollowing day the dehydration was continued and the cell monolay-ers were embedded in EPON. Thin parallel sections, 40-50 nm, of theembedded cell monolayers were cut using a Reichert microtome.Sections were post-stained with 4% uranyl acetate for 5 minutesfollowed by 1% lead citrate for 3 minutes, and observed in a Philips400 electron microscope.

Cell permeabilization experimentsFor these experiments the method of Symons and Mitchison (1991)

was used with some minor modifications. Briefly, HeLa cells weregrown on uncoated 11 mm glass coverslips for 24 hours beforeinfection with virus. At 8 hours p.i. the cells were transferred to obser-vation medium (OM; bicarbonate free MEM buffered with 20 mMHepes, pH 7.4, 10% FCS, penicillin and streptomycin) at 37°C.Before the experiment the cells were cooled to room temperature andthe subsequent experiments carried out at this temperature. Thecoverslip was incubated cell side down on two drops of rinsing buffer(RB; 20 mM Hepes, pH 7.5, 138 mM KCl, 4 mM MgCl2 and 3 mMEGTA) and and then permeabilized for 30 seconds on a drop of incu-bation buffer (IB; RB plus 0.2 mg/ml saponin). A 10 mg/ml stocksolution of rhodamine G-actin (Cytoskeleton, USA) stored at −80°Cwas pre-thawed, diluted 1:10 with G-actin buffer (2 mM Tris-HCl,pH 8.0, 0.1 mM ATP, 0.1 mM DTT and 0.1 mM CaCl2) and clarifiedfor 20 minutes at 26 psi in an airfuge at 4°C. It was then diluted to afinal concentration of 1.0 or 0.3 µM in IB plus 1 mM ATP. The cov-erslips were incubated on 20 µl drops of rhodamine actin for varioustimes from 5 seconds to 10 minutes. After incubation in rhodamineactin the cells were immediately fixed in 3% PFA in 10 mM MES,150 mM NaCl, 5 mM EGTA, 5 mM MgCl2 and 5 mM glucose (CB)for 10 minutes. The coverslips were then permeablized in 0.1% TX-100 for 1 minute and further fluorescence labelling carried out asdescribed above. All permeabilization incubation experiments wereexamined and recorded using the EMBL confocal microscope facility.The resulting digital images were converted into false colour, mergedand annotated using the Adobe 3.0 software package.

RESULTS

The structure of vaccinia actin tailsCells stained with bodipy phalloidin 6 hours or later afterinfection with vaccinia virus show numerous actin tails andprojections (Fig. 1A) that are not seen in uninfected cells. Thenumber of actin tails and projections in each cell is variable,between 1-200 in the same infection experiment, but theirnumber increases with longer infection times. When virusparticles extend from the plasma membrane the projectionsinitially have a broad and constant diameter along their length(Fig. 1B). However, as the projections continue to extend, theregion closer to the cell appears withered and thin (Fig. 1B).

Closer inspection of the appearance of vaccinia-inducedactin tails shows that although similar to those of bacterialpathogens, there are a number of differences in their organiz-ation. Rather than being smooth like those in the publishedimmunofluorescence images of bacterial comets, the vacciniaactin tails have a somewhat splayed appearance with branch-like appendages extending outward randomly along theirlength (Fig. 2). These branches project out at an angle of about45° with respect to the central axis of the tail, towards the virusparticle end. This splayed appearance along the completelength is seen mainly in intracellular tails. Microvillar projec-tions at the cell surface tend to be smoother with only the tipssplaying outwards. In cases where projection tips do not splay,the region of actin in contact with the virus particle shows aclear cup-like structure (Fig. 1B). The same cup-like structureis seen in intracellular tails but is less pronounced (data notshown).

The general overall appearance of the actin tail in theelectron microscope, after myosin S1 decoration of the actinfilaments, is in good agreement with those in immunofluores-cence microscopy (Figs 3 and 4). In electronmicrographs actinfilaments in the tail splay outwards, as suggested by the

1741Vaccinia virus and membrane-actin interactions

Fig. 1. Immunofluorescence images ofvaccinia infected cells. (A) The overallappearance of the actin cytoskeleton,visualized by phalloidin staining, of aHeLa cell 8 hours after vaccinia infection.(B) A long projection from the cellsurface with a withered base. Such longprojections are often observed to be over20 µm in length. In addition, shorterprojections with a cup-shaped appearanceat the viral end are visible (arrowheads).Bars, 25 µm.

branches seen in fluorescence microsopy. This is especiallyevident in curled intracellular tails (Fig. 3). Thus the tailappears to be arranged almost like an ‘upside-down Christmastree’ with the filaments oriented so that they extend outwardsfrom the centre of the tail. Actin filaments are clearly seen incontact with the membrane around the virus particle (Fig. 3).In addition, when virus particles are projecting from the cellsurface, a number of actin filaments from the tail are seen incontact with the plasma membrane both in longitudinal andtransverse sections (Fig. 4). In longitudinal sections a numberof actin filaments are seen to interact with the plasmamembrane with their fast-growing ends (Fig. 4). Electronmicrographs of long projections with withered bases clearlyshow that actin filaments are still present along the wholelength of the projection, although at a much reduced densitynear the cell where they appear to be arranged in a much moreparallel fashion (Fig. 4).

Polarity and length of actin filaments in the tailSuperposition of filaments over each other in the same sectionwithin the bulk of the tail viewed by electron microscopy

Fig. 2. High resolution immunofluorescence images of intracellulartails visualized with phalloidin. Both panels show branch likestructures splaying outwards from tapering intracellular tails.Arrowheads indicate the position of virus particles at the broader endof the tails. Bar, 2 µm.

makes polarity determination by myosin S1 decorationdifficult. However, while we cannot determine the polarity inthe majority of cases, it is possible to discern the characteris-tic myosin S1 arrowhead polarity on a number of filamentstowards or at the edge of the actin tails in very thin sections(40-50 nm). In these cases the barbed or fast growing end ofthe actin filaments point towards the virus (Fig. 4). As withfilament polarity determination, it is difficult to measure thetrue length of individual actin filaments within the tail. In ourelectronmicrographs we find filaments of up to 0.74 µm in pro-jections (Fig. 4), although this is probably an underestimationof the true length.

Actin recruitment into intracellular tails occurs fromthe virus surface in a polarized fashionThe fact that the barbed ends of the filaments point towards thevirus particle suggests that actin monomers incorporate intotails at or near the virus surface. To investigate if this wasindeed the case, we adopted the technique of Symons andMitchison (1991), in which sites of actin polymerization wereobserved by the incorporation of rhodamine labelled G-actininto cells which were permeabilized with saponin. Phalloidinstaining of uninfected permeablized cells confirmed that underthe conditions of our assay, cells still maintained a highlyorganized actin cytoskeleton that incorporated rhodamine actinat focal adhesions, even after 10 minutes of permeablization(Fig. 5).

A 10 second incubation of permeabilized infected cells with1.0 µM rhodamine G-actin resulted in incorporation of actinonly at virus particles which had associated actin tails, but notinto the tails themselves which were visualized with phalloidin,i.e. total F-actin (Fig. 6A-C). Only a proportion of virusparticles recruited rhodamine actin, which is consistent withthe number of virus particles which induce tails (Cudmore etal., 1995). There was no apparent difference in the ability ofvirus particles to recruit rhodamine actin, whether they wereon intracellular actin tails or extending outwards from theplasma membrane on actin projections (Fig. 6). When thelength of incubation time with rhodamine G-actin wasincreased, incorporation of actin into the intracellular tails onlyoccurred from the virus particle which remained covered in abright cloud of G-actin (Fig. 6). No rhodamine actin incor-poration was observed along the tail that did not originate fromthe virus particle, irrespective of incubation time. This obser-

1742 S. Cudmore and others

Fig. 3. Electron microscopy of S1 myosindecorated intracellular actin tails. The mainpanel shows a highly curled intracellular tailwith the characteristic splaying of actin filamentsoutwards at an angle of ~45° to the central axisof the tail. The smaller inset shows anintracellular tail with numerous actin filaments incontact with the virus particle. Note that thevirus particles are oriented with their longer axesperpendicular to the long axis of the tail. Thestars indicate virus particles. Bar, 400 nm.

vation suggests that the branch-like structures which extendalong the length of intracellular tails (see Figs 2-4) do not formby the addition of actin monomers at their ends as they do notincorporate rhodamine actin from their tips. Experimentsperformed with 0.3 µM rhodamine actin, a value below thecritical concentration for assembly of actin at the pointed end,0.7-0.9 µM in vitro, but above the value of 0.1 µM for barbedend assembly, gave identical results.

Antibody labelling of virus particles together withrhodamine actin incorporation shows that at both early and lateincubation time-points actin recruitment onto the viral surfaceoccurs in a polarized fashion immediately adjacent to the tail(Fig. 6).

Vaccinia-actin nucleation sites are transferred to theplasma membraneAs Fig. 6 shows, the rhodamine actin incorporating structure inthe virally induced projections, i.e. the actin nucleating surfaceon the virus particle, tends to be cup-shaped. However, in manycases these nucleating structures had fragmented into severalfoci, which were still able to nucleate actin polymerization (Fig.7). We also occasionally saw projections that actively incorpo-rated actin into these foci although there was no virus particleat the tip (data not shown). The simplest explaination for ourobservations is that the actin nucleation sites are now present

in the plasma membrane as a result of viral membrane fusionand are free to diffuse within the plasma membrane.

DISCUSSION

Little is known about the interactions between actin filamentsand membranes. Previous observations demonstrated thatvaccinia virus is capable of inducing the formation of actin-rich microvilli during its infection cycle (Stokes, 1976; Hilleret al., 1979, 1981; Krempien et al., 1981). We have recentlyextended these earlier observations to show that the IEV formof vaccinia induces the formation of actin tails that are similarto the actin tails seen in cells infected with bacterial pathogenssuch as Shigella, Listeria, and Rickettsia (Tilney and Tilney,1993; Cossart and Kocks, 1994; Cossart, 1995). The IEV formof vaccinia virus is surrounded by a membrane cisterna derivedfrom the trans-Golgi network (Schmelz et al., 1994) and con-sequently it offers a unique opportunity to examine membrane-actin interactions in detail. As a first step towards a detailedcharacterization of the system, we have examined the structureof the actin tails induced by IEV.

The ultrastructure of the vaccinia actin tailElectron micrographs of both intracellular tails and

1743Vaccinia virus and membrane-actin interactions

Fig. 4. Electron microscopy of virally-induced projections. (A) A virus particle (*), which appears to be outside the plasma membrane, with itslonger axis perpendicular to the projection. In some cases the polarity of the S1 myosin decorated actin filaments can be discerned, with thebarbed or fast-growing end of the filaments oriented towards the virus particle (arrowheads). Interactions between individual actin filamentsand the plasma membrane are indicated by arrows. The length of a single filament is indicated. (B) A long projection that lacks a virus particleat its tip and shows a reduced density of actin filaments at its withered base. In addition, the filaments in the base tend to have a more parallelarrangement than at the top of the projection. Arrows indicate actin filament-membrane interactions and the length of a single filament isindicated. The inset in B shows a transverse section through the top of a projection. The density distribution of actin filaments is consistent withlongitudinal sections. Arrows indicate actin filament-membrane interactions. Bars, 400 nm.

microvilli-like projections, show many features that are con-sistent with the immunofluorescence images, suggesting thatour fixation and S1 labelling process has not drasticallyaltered the structure of the tail. It is clear that actin filamentsin the vaccinia tail are arranged in an extremely densemeshwork, at least as dense as Listeria tails (compare Figs 3and 4 with Figs 16 and 17 in Tilney and Portnoy, 1989, andFig. 4B in Kocks et al., 1992). Another common featurebetween the tails of vaccinia and Listeria is actin filamentpolarity. Although we cannot determine actin filamentpolarity in the bulk of the viral tail, both systems show thatactin filaments at the edge of the tail have unidirectionalpolarity, with their barbed or fast-growing ends in thedirection of motion. Although similar in their polarity, wefind that vaccinia tails contain longer actin filaments than the

0.2 µm length filaments reported for Listeria systems (Tilneyand Portnoy, 1989) but more in agreement with those seen inRickettsia induced actin tails (Heinzen et al., 1993). Giventhe severe limitations for measurement from sections of actintails, we suspect that the filaments may be significantlylonger.

The most striking difference between the two systems isseen in the organization of actin filaments in the tail. Actinfilaments in vaccinia tails are oriented so that they splayoutwards at a 45° angle from the central axis of the tail. Thisis not the case in Listeria tails where actin filaments inelectron micrographs appear to be more randomly oriented(Tilney and Portnoy, 1989). In addition, fluorescence polar-ization microscopy shows that while filaments in the core ofthe Listeria tail are randomly oriented, those at the surface or

1744 S. Cudmore and others

Fig. 5. Incorporation of rhodamine actininto permeabilized HeLa cells identifiessites of actin nucleation andpolymerization. (A) The complete actincytoskeleton visualized with biodipyphalloidin; (B) localization ofrhodamine actin incorporation sites aftera 10 minute incubation with rhodamineactin, and (C) merged false colour

images of phalloidin (green) and rhodamine actin images (red) showing that although focal adhesions have incorporated rhodamine actin,they also maintain a pool of unincorporated G-actin at their tips. Bar, 10 µm.

Fig. 6. Incorporation of rhodamine actininto vaccinia actin tails occurs from theviral particle surface in an asymmetricfashion. (A,D,G) Total F-actin visualizedby phalloidin; (B,E,H) localization ofrhodamine actin after a 10 second, 1minute and 3 minute incubation,respectively. (C,F,I) Merged false colourimages of the phalloidin signal (green)and the rhodamine signal (red) for A,B,D,E, and G,H, respectively. From theseimages, it is clear that viral particlesrecruit rhodamine G-actin to theirsurface which then becomesincorporated into the tail. After only 10seconds virus particles are labelled withrhodamine G-actin but after 1 minute therhodamine signal is seen about halfwaydown the tail when compared to thephalloidin signal. After 3 minutes,rhodamine actin has incorporated downthe complete length of the tails as judgedby the rhodamine and phalloidin signals.In all merged panels it is noticeable thatthere is not 100% co-localizationbetween the rhodamine signal on thevirus particle and the phalloidin signal ofthe tail, suggesting the virus maintains apool of G-actin on its surface. (J) Virusparticles. (K) Rhodamine actinincorporation after 10 seconds.Arrowheads in the merged false colourimage in L highlight that rhodamineactin incorporation occurs in anpolarized fashion on the virus particlesurface. Bar, 4 µm.

shell of the tail are preferentially aligned parallel to thedirection of movement (Zhukarev et al., 1995). This differ-ence in structural organization of actin filaments probably

reflects a difference between the two systems in both thenature of the actin nucleation site, as well as the actin cross-linking proteins in the tail.

1745Vaccinia virus and membrane-actin interactions

Fig. 7. Actin nucleation sites at theprojection tips fragment into foci. (A andD) Rhodamine actin incorporation after 3minutes. (B) Total F-actin visualized withbiodipy phalloidin. (E) Labelling of virusparticles. (C and F) Merged false colourimages of the rhodamine signal (red) andphalloidin or virus signal signal (green)for A,B and D,E, respectively.Arrowheads in A to C indicate a branchnear the tip of a projection incorporatingrhodamine actin i.e. nucleating actinpolymerization. Arrowheads in D to Fhighlight the fragmentation of the actinnucleating surface at the tip of aprojection into discrete foci, many ofwhich are not in contact with the virusparticle. Bar, 3 µm.

Fig. 8. Model depicting the development of virally inducedmicrovilli and fusion of the outer membrane of the IEV with theplasma membrane at the tip of this projection. The plasma membraneis shown in green and actin filaments in the projection are blue. TheIEV is shown in black with its outer most membrane depicted in red.In A, the actin nucleation sites present in the viral membranenucleate actin polymerization and consequently the projectionextends outwards. (B) The outermost membrane of the IEV form ofvaccinia (red), along with its actin nucleation sites, fuses with theplasma membrane, releasing the CEV outside the projection ready toinfect neighbouring cells upon contact. (C) The actin nucleationsites, which are now present in the plasma membrane, begin todiffuse but still continue to nucleate actin filaments as the projectionextends.

Nucleation of actin tails occurs asymmetrically atthe viral surfaceUnder the conditions of our permeabilization assay the actincytoskeleton of uninfected cells remains intact and readilyincorporates rhodamine actin at focal adhesions. However,video analyses of permeablized infected cells shows that virustipped projections do not extend; we could not clearly distin-guish intracellular tails due to the loss of phase contrast as aresult of the detergent treatment (data not shown). Thus, therates of rhodamine actin incorporation in infected perme-ablized cells are independent of virus movement. However, ourrhodamine actin incorporation experiments still specificallyidentify localized sites of actin nucleation within the cell.

From our results it is clear that the first stage of tailformation is the recruitment of G-actin, and not pre-formedactin filaments, to the virus particle. Furthermore, continuedtail growth requires a constant recruitment of G-actin as viralparticles remain coated in rhodamine actin at all stages ofincorporation experiments. We also observe that actin nucle-ation occurs in an asymmetric fashion on the virus particlesurface, as is also observed for Listeria (Kocks et al., 1993)and has been shown to be a prerequisite for Listeria movement(Smith et al., 1995). We are currently unable to explain whyactin recruitment is polarized on the virus particle surface,given that the virus particle is covered in a membrane that hasno obvious asymmetry. However, the virus particle is asym-metric in shape and in our electron micrographs they are nearlyalways oriented with their long axes perpendicular to the tail(Fig. 3).

Incorporation of rhodamine actin into the tail only occursfrom the virus surface and not from internal sites along the tail.Thus, actin branches on intracellular tails do not incorporatelabel from their tips. This is an important observation as itsuggests that the only free ends available for polymerization ofactin monomers are found at, or near, the viral surface and thatfilaments within the tail or branches are blocked from furtheraddition of actin. Given that the virus particle at the tipnucleates actin tail assembly, a constant supply of actin

monomers must be provided at the projection tip even whenextentions are more than 20 µm long. This pool of actin mustcome from either the cell or alternatively may be supplied fromfilaments depolymerizing at the base of the tail. We also findthat there is no difference in the incorporation of rhodamineactin, regardless of whether it is used above or below thecritical concentration for assembly of actin at the pointed endin vitro. Hence polymerization only occurs at the free barbed

1746 S. Cudmore and others

ends found at the virus particle surface, in agreement withfilament polarity.

Transfer of viral-actin nucleation sites to the plasmamembraneOne feature that was noticeable in long projections emergingfrom the plasma membrane was that the tips of projectionswere often fragmented into several foci (Fig. 7). These fociwere still capable of incorporating rhodamine actin. We wouldlike to suggest that these foci originate when the outermembrane of the IEV fuses with the plasma membrane torelease the second infectious form. Such a fusion event wouldtransfer actin nucleation sites from the membrane around theIEV to the plasma membrane where nucleation appears tocontinue despite the fact that the particle is now localized onthe outerside of the plasma membrane (Fig. 8). In support ofthis hypothesis we often see virus particles which appear to beon the outside of microvillar projections, suggesting that theyhave already fused with the plasma membrane (Fig. 4). Thatthese particles are indeed EEV can be confirmed by conven-tional thin section electron microscopy analysis under con-ditions that avoid cell permeabilization (data not shown). Asimilar conclusion can be drawn from images in the originalhigh voltage microscopy study of vaccinia virus release byStokes (1976). Consistent with this, it has previously beenreported that the IEV is released from the cell by fusion of itsouter membrane with the plasma membrane, releasing the EEV(Dales, 1971; Ichihashi et al., 1971; Morgan, 1976; Blasco andMoss, 1991, 1992). However, some of the EEV can remainattached to the outer surface of the cell, these being referred toas CEV (Blasco and Moss, 1991, 1992). It has been proposedthat while the EEV is responsible for long range spread ofvirions (Appleyard et al., 1971; Boulter and Appleyard, 1973;Payne, 1980), the role of the CEV is in direct cell-to-cell spread(Blasco and Moss, 1992). We believe that the virus particleson the outer suface of projection tips are CEV and thus actinpolymerzation provides the mechanism by which theseparticles contact neighbouring cells in order to cause the directcell to cell spread of infection.

ConclusionsGiven the ease with which the vaccinia virus sytem can bemanipulated, we believe that vaccinia virus offers a uniqueopportunity to understand the interactions between actinfilaments and membranes at a molecular level. Production ofrecombinant vaccinia virus is well established offering the pos-sibility of generating deletion strains to dissect the viral factorsinvolved in the motile process. In addition, through the use ofthe recombinant T7 vaccinia virus expression system it ispossible to couple vaccinia infection with overexpression ofvarious proteins, including actin binding proteins, to manipu-late actin cytoskeleton dynamics during infection. We havealready observed that the recombinant T7 vaccinia virusexpression system induces actin tails that are indistinguishablefrom wild-type strains (data not shown). Furthermore, there aremany striking similarities between the actin filamentsnucleated from the membrane surrounding vaccinia virus andthose at the leading edge of the cell. In both situations, the actinfilaments are organised in an orthogonal network at an angleof 45° with the barbed ends of the filaments pointing towardsthe nucleating membrane (Small et al., 1995). We therefore

believe that analysis of the vaccinia virus system will greatlyfacilitate understanding of actin-membrane interactionsoccurring at the leading edge of motile cells.

We thank Drs Ernst Steltzer and Sigrid Reinsch for advice con-cerning confocal microscopy, Dr Michael Glotzer for advice onimaging, Dr Daniella Klaus for a gift of S1 myosin and Dr RafaelBlasco for interesting discussions on virus-membrane fusion. We alsothank Cytoskeleton, USA for rhodamine-actin and Drs AntonyHyman, Rebecca Heald and Sigrid Reinsch for critical reading of themanuscript as well as the referees for improving the paper.

REFERENCES

Appleyard, G., Hapel, A. J. and Boulter, E. A. (1971). An antigenicdifference between intracellular and extracellular rabbitpox virus. J. Gen.Virol. 13, 9-17.

Blasco, R. and Moss, B. (1991). Extracellular vaccinia virus formation andcell-to-cell virus transmission are prevented by deletion of the gene encodingthe 37,000-dalton outer envelope protein. J. Virol. 65, 5910-5920.

Blasco, R. and Moss, B. (1992). Role of cell-associated enveloped vacciniavirus in cell-to-cell spread. J. Virol. 66, 4170-4179.

Bohn, W., Rutter, G., Hohenberg, H., Mannweiler, K. and Nobis, P. (1986).Involvement of actin filaments in budding of measles virus: studies oncytoskeletons of infected cells. Virology 149, 91-106.

Boulter, E. A. and Appleyard, G. (1973). Differences between extracellularand intracellular forms of poxvirus and their implications. Progr. Med. Virol.16, 86-108.

Cossart, P. and Kocks, C. (1994). The actin-based motility of the faculativeintracellular pathogen Listeria monocytogenes. Mol. Microbiol. 13, 395-402.

Cossart, P. (1995). Actin based bacterial motility. Curr. Opin. Cell Biol. 7, 94-101.

Cudmore, S., Cossart, P., Griffiths, G. and Way, M. (1995). Actin-basedmotility of vaccinia virus. Nature 378, 636-638.

Dales, S. (1971). Involvement of membranes in the infectious cycle of vaccinia.In Cell Membranes, Biological and Pathological Aspects (ed. G. W. Richterand D. G. Scarpelli), pp. 136-144. Williams & Wilkins, Baltimore.

Doms, R. W., Blumenthal, R. and Moss, B. (1990). Fusion of intracellular-and extracellular forms of vaccinia virus with the cell membrane. J. Virol. 64,4884-4892.

Giuffre, R., M., D. Tovell, R., C. Kay, M. and Tyrell, D. L. J. (1982).Evidence for an interaction between the membrane protein of aparamyxovirus and actin. J. Virol. 42, 963-968.

Heinzen, R. A., Hayes, S. F., Peacock, M. G. and Hackstead, T. (1993).Directional actin polymerization associated with spotted fever groupRickettsia infection of Vero cells. Infect. Immun. 61, 1926-1935.

Hiller, G., Weber, K., Schneider, L., Parajsz, C. and Jungwirth, C. (1979).Interaction of assembled progeny pox viruses with the cellular cytoskeleton.Virology 98, 142-153.

Hiller, G., Jungwirth, C. and Weber, K. (1981). Fluorescence microscopicalanalysis of the life cycle of vaccinia virus in the chick embryo fibroblasts.Exp. Cell Res. 132, 81-87.

Ichihashi, Y., Matsumoto, S. and Dales, S. (1971). Biogenesis of poxviruses:Role of A-type inclusions and host cell membranes in virus dissemination.Virology 46, 507-532.

Kocks, C., Gouin, E., Tabouret, M., Berche, P., Ohayon, H. and Cossart, P.(1992). L. monocytogenes-induced actin assembly requires the actA geneproduct, a surface protein. Cell 68, 521-531.

Kocks, C., Hellio, R., Gounon, P., Ohayon, H. and Cossart, P. (1993).Polarized distribution of Listeria monocytogenes surface protein ActA at thesite of directional actin assembly. J. Cell Sci. 105, 699-710.

Krempien, U., Schneider, L., Hiller, G., Weber, K., Katz, E. andJungwirth, C. (1981). Conditions for pox virus specific microvilli formationstudied during synchronized virus assembly. Virology 113, 556-564.

Morgan, C. (1976). Vaccinia virus reexamined: Development and release.Virology 73, 43-58.

Murti, K. G., Chen, M. and Goorha, R. (1985). Interaction of frog virus 3with the cytomatrix. III. Role of microfilaments in virus release. Virology142, 317-325.

Payne, L. G. (1980). Significance of extracellular enveloped virus in the invitro and in vivo dissemination of vaccinia. J. Gen. Virol. 50, 89-100.

1747Vaccinia virus and membrane-actin interactions

Payne, L. G. and Kristensson, K. (1985). Extracellular release of envelopedvaccinia virus from mouse nasal epithelial cells in vivo. J. Gen. Virol. 66,643-646.

Rodriguez, J. F., Janeczko, R. and Esteban, M. (1985). Isolation andcharacterization of neutralizing monoclonal antibodies to vaccinia virus. J.Virol. 56, 482-488.

Schmelz, M., Sodeik, B., Ericsson, M., Wolffe, E. J., Shida, H., Hiller, G.and Griffiths, G. (1994). Assembly of vaccinia virus: The second wrappingcisterna is derived from the trans Golgi network. J. Virol. 68, 130-147.

Small, J. V., Herzog, M. and Anderson, K. (1995). Actin filamentorganization in the fish keratocyte lamellipodium. J. Cell Biol. 129, 1275-1286.

Smith, G., A., D. Portnoy, A. and Theriot, J. A. (1995). Asymmetricdistribution of the Listeria monocytogenes ActA protein is required andsufficient to direct actin-based motility. Mol. Microbiol. 17, 945-951.

Sodeik, B., Doms, R. W., Ericsson, M., Hiller, G., Machamer, C. E., Hof,W. v. t., Meer, G. v., Moss, B. and Griffiths, G. (1993). Assembly ofvaccinia virus: Role of the intermediate compartment between theendoplasmic reticulum and the Golgi stacks. J. Cell Biol. 121, 521-541.

Stokes, G. V. (1976). High-voltage electron microscope study of the release ofvaccinia virus from whole cells. J. Virol. 18, 636-643.

Symons, M. H. and Mitchison, T. J. (1991). Control of actin polymerization inlive and permeabilized fibroblasts. J. Cell Biol. 114, 503-513.

Tilney, L. G. and Portnoy, D. A. (1989). Actin filaments and thegrowth,movement and spread of the intracellular bacterial parasite, Listeriamonocytogenes. J. Cell Biol. 109, 1597-1608.

Tilney, L. G. and Tilney, M. S. (1993). The wily ways of a parasite: inductionof actin assembly by Listeria. Trends Microbiol. 1, 25-31.

Tyrell, D. L. and Norrby, E. (1978). Structural polypeptides of measles virus.J. Gen. Virol. 39, 219-229.

Zhukarev, V., Ashton, F., Sanger, J. M., Sanger, J. W. and Shuman, H.(1995). Organization and structure of actin filament bundles in Listeriainfected cells. Cell Motil. Cytoskel. 30, 229-246.

(Received 1 March 1996 - Accepted 22 April 1996)