The Tailless Icosahedral Membrane Virus PRD1 Localizes the Proteins Involved in Genome Packaging and Injection at a Unique Vertex

The double-stranded DNA (dsDNA) virus PRD1 carries its genomein a membrane surrounded by an icosahedral protein shell. Theshell contains 240 copies of the trimeric P3 protein arrangedwith a pseudo T = 25 triangulation that is reminiscent of themammalian adenovirus. DNA packaging and infection are believedto occur through the vertices of the particle. We have usedimmunolabeling to define the distribution of proteins on thevirion surface. Antibodies to protein P3 labeled the entiresurface of the virus. Most of the 12 vertices labeled with antibodiesdirected against proteins P5, P2, and P31. These proteins areknown to function in virus binding to the cell surface. ProteinsP6, P11, and P20 were found on a single vertex per virion. TheP6 and P20 proteins are believed to function in DNA packaging.Protein P11 is a pilot protein that is involved in a complexthat mediates the early stages of DNA entry to the host cell.Labeling with antibodies to P5 or P2 did not affect the labelingof P6, the unique vertex protein. Labeling with antibodies tothe unique vertex protein P6 interfered with the labeling byantibodies to the unique vertex protein P20. We conclude thatPRD1 utilizes 11 of its vertices for initial receptor binding.It utilizes a single, unique vertex for both DNA packing duringassembly and DNA delivery during infection.

INTRODUCTION

Top

Introduction

A crucial step in virus assembly is the recognition of virus-specificnucleic acid and its consequent encapsidation into the virusparticle. Two principal mechanisms have been described: cocondensationof the viral genome with the assembling coat and formation ofan empty polyhedral procapsid into which the virus genome ispackaged.

The first mechanism was initially illustrated using tobaccomosaic virus (16) and is assumed to occur during the assemblyof small RNA nonmembranous and membrane-containing viruses (15,28, 30). The second mechanism is well described for tailed double-strandedDNA (dsDNA) bacteriophages (11, 37). It may be a common mechanismof genome packaging for complex icosahedral viruses such asherpes simplex virus (39).

Viral capsids of the second type are not simply static compartmentsthat protect the genome but dynamic molecular machines thattranslocate the viral nucleic acid into the capsid so that thepacking nears crystalline density. One of the icosahedral capsidvertices (31) is specialized for nucleic acid translocation.This portal vertex (54) contains a multimeric portal proteinor connector, since it connects the phage head to the tail.The portal typically has a different symmetry than the vertex,which must respect the fivefold symmetry of its icosahedralposition. The connectors of some tailed dsDNA phages are dodecamers(50) on the virus. The symmetry mismatch between the fivefoldvertex and the packaging protein is believed to facilitate movement(50) or rotation (27) during genome translocation through thevertex. A separate ATPase (the terminase [17]) that is not partof the capsid is found in dsDNA phages. The terminase recognizesand binds to the phage DNA, associates with the portal complex,and provides the energy for the translocation event (17). Asymmetry mismatch is often found at the site of packaging. Forexample, a hexameric protein, P4, has been found at the verticesof a membrane dsRNA virus, ${phi}$ 6 (18, 29; E. J. Mancini, D. H. Bamford,J. M. Grimes, and D. I. Stuart, unpublished data). The ${phi}$ 6 portalprotein is an NTPase that provides the energy for translocation(41, 42).

The membrane-containing bacteriophage PRD1 has a linear 15-kbpgenome enclosed in a membrane vesicle that is surrounded bya protein coat (Fig. 1). The genomic DNA packaging is believedto occur through a vertex, although this has not been directlydemonstrated. The vertices are also responsible for receptorrecognition (23, 44) through a spike-like structure containingproteins P2, P5, and P31. The vertices are also assumed to bethe site for the initiation of DNA delivery through a complexinvolving the pilot protein P11 (53a). The combination of structuralinformation obtained by cryo-electron microscopy (13, 46) andimage processing with that obtained by X-ray crystallography(9, 10, 53) shows a surprising similarity to that obtained foradenovirus (1, 51, 52). This led to a proposal that these virusesshare a common ancestor and that viruses may form lineages thathave members infecting hosts in all domains of life (3).

View larger version (159K):
[in this window]
[in a new window]
FIG. 1. Labeling of wt PRD1 with antibodies to P3, P5, and P2. PRD1 particles were labeled with antibodies and visualized using 10-nm gold coupled to protein A as described in the Materials and Methods. (A) The extensive labeling of wt PRD1with anti-P3 is shown. (B) Labeling of a mixture of wt PRD1 and MMTV (arrows) with the same anti-P3 is shown. No labeling of MMTV is visible. (C and D) Labeling of a PRD1/MMTV mixture with antibodies against P5 (C) and P2 (D) is also shown. The labeling, while less extensive than that with P3, was associated with the PRD1 and not with the MMTV particles (arrows).

The order of PRD1 assembly is unusual for a membrane-containingvirus. PRD1 has an assembly intermediate that contains the capsidand the membrane but no DNA (34). These particles package DNAand mature to virions. This order is more similar to that describedfor tailed dsDNA phages than to that described for other membraneviruses (2, 30). The behavior of the packaging-deficient mutant,sus1, provides support for this idea. This mutant has a defectin protein P9 that contains an ATP-bridging Walker motif (7).The mutation leads to the formation of membrane-containing particlesthat lack DNA. Two other mutant classes, lacking the integralmembrane proteins P20 and P22, also show packaging-associateddefects (33, 34). Intracellular particles of both mutants packageDNA but release it upon particle isolation. Genetic evidencelinks protein P6 to a complex involved in packaging (53a). Attemptsto demonstrate a special packaging vertex by structural methodshave been less effective (47). The use of icosahedral symmetry(2, 14, 18, 20) in averaging blurs the density of the differing1 in 12 vertices.

PRD1 injects DNA into its gram-negative hosts by a complex multistepmechanism. The vertex-associated spike complex, containing P2,P5, and P31, binds to a cell surface receptor via the elongatedP2 protein. This interaction triggers a cascade of events thatlead to the penetration of the cell envelope by a virally derivedmembrane tube (24, 25) and delivery of the DNA into the cytosol(24). This process involves the viral proteins P7/P14 and P11,P15, P16, P18, and P32 (see references in reference 17) thatinteract with the host membranes and digest the peptidoglycanlayer. The viral membrane is involved in the final DNA translocationprocess, and empty particles frequently display a tubular structureprotruding from a vertex (4). The initial binding of P2 is reversible;however, the interaction that leads to penetration of the hostcell wall is not. It had been proposed that any vertex can functionin DNA delivery, since the receptor binding spike is found atthe vertices and it is logical to associate binding with penetration(4, 24, 25, 32, 43, 44). This assumption has never been testeddirectly. Infection in other membrane-containing viruses isa multistep process that does not proceed directly from bindingto productive infection (19, 30, 48, 49, 56).

We can test the possibility of separate functions and structuresfor the different vertices by forgoing the use of icosahedralaveraging that would mask the differences between them. Theavailability of poly- and monoclonal antibodies (8, 23, 26,44) against several PRD1 structural proteins allows an immunoelectronmicroscopy approach to the structure. The high contrast affordedby gold labeling allows differentiation of the composition ofthe sites, although the detailed structural information is lost.We have identified two types of vertices in PRD1. The first,or initial binding, vertex contained the receptor binding P2protein as well as the associated vertex components P31 andP5. The second, or translocation, vertex contained proteinsP6 (53a) and P20, which are involved in genome packaging, andprotein P11, which serves as a pilot protein for the formationof the tube that conducts the genome into the target cell. Only1 translocation vertex and 11 binding vertices were presentper virion. Our labeling studies showed that these verticeswere distinct; the translocation vertex did not contain thebinding vertex components. Infection by PRD1 must thereforeinvolve a two-step process similar to that in other membraneviruses in which reversible binding at multiple vertices leadsto permanent attachment and tube formation at the unique translocationvertex.

MATERIALS AND METHODS

Top

Materials and Methods

Bacteria and phages. Bacterial cells were grown in Luria-Bertani medium (45). Wild-type(wt) PRD1 was propagated on Salmonella enterica serovar Typhimuriumstrain DS88 (6). sus539 is an amber mutant with a mutation ingene II encoding the receptor binding protein P2. Growth innonsuppressor strain DS88 produces particles that do not haveprotein P2 but have all the other viral proteins that can bedetected (23). These particles are thus described as P2^-. sus690is an amber mutant with a mutation in gene V. The suppressorhosts were DB7156(pLM2) for sus539 and PSA(pLM2) for sus690(33, 35). Particles lacking both proteins P5 and P2 (P2^- andP5^-) are produced when sus690 is propagated in the nonsuppressorstrain DS88 host. This phenomenon does not reflect genetic polarity,since protein P2 is produced in normal amounts. It representsan assembly defect. The receptor binding protein P2 is associatedwith the viral particle via protein P5 (6).

All viruses were grown with strain DS88 and purified on 5 to20% linear sucrose gradient (5) for particle production. Thevirus zone from the gradient was either collected by differentialcentrifugation (Sorval T865 rotor; 32,000 rpm, 3 h, +5°C)or applied for use in ion exchange chromatography (Sartorius).The Sus539 particles were purified by sucrose gradient centrifugationfollowed by ion exchange chromatography on a MemSep filter andelution with 5 mM potassium phosphate (pH 7.2)-300 to 400 mMNaCl (55).

Protein concentrations of the virus preparations were determinedusing bovine serum albumin as a standard (12). Sodium dodecylsulfate-polyacrylamide gel electrophoresis was performed asdescribed previously (40). The proteins were either stainedwith Coomassie brilliant blue or transferred from the gel ontoa polyvinylidene difluoride membrane (Millipore) and detectedwith antibodies (Western blotting).

Antibodies. Specific rabbit antisera recognizing phage PRD1 proteins P2(23), P5 (26), and P31 (8, 44), as well as monoclonal antibodies6A488 (recognizing protein P6), 7N41 (recognizing protein P7/P14),11A401 (recognizing protein P11), and 16T56 (recognizing proteinP16) (26), were used for labeling. The polyclonal serum againstP20 was raised using a specific peptide as an antigen. The peptideCNHFGKNAVTSE (linked to the keyhole limpet hemocyanin via thecysteine residue) was purchased from KJ Ross-Petersen AS, Holte,Denmark. For the antiserum production, a rabbit was immunizedthree times at 3-week intervals with 900 µg of the antigenemulsified with either Freund's complete (for the first immunization)or incomplete adjuvant. The chemiluminescent detection of peroxidase-conjugatedsecondary antibodies was performed using either an ECL Plus(Amersham) or a Super Signal R West Pico (Pierce) system.

Antibody labeling. All steps were performed in a humidified room-temperature chamberwith Parafilm at its base (21, 22, 57). Glow-discharged, carbon-coatedFormvar 300-mesh nickel grids were placed onto 10- to 15-µlvirus aliquots for 1 min. All subsequent steps were performedby floating the grid with the specimen side down on aliquotsof the solutions (25 to 30 µl). After two 1-min washesin phosphate-buffered saline (PBS) the grids were fixed in 0.5%glutaraldehyde in PBS for 15 min. After three 5-min washes inPBS, the grids were treated with 0.5% Triton X-100 in PBS for15 min. After four 5-min washes in PBS, the potential residual-freeglutaraldehyde was neutralized for 15 min in PBS containing0.02 M glycine. After three 5-min washes in PBS, the grids werethen treated with PBS containing 1.0% ovalbumin (albumin, chickenegg, catalogue no. A-5503; Sigma) for 10 min. The grids werethen exposed for 1 h to the primary antibodies diluted in PBS-1.0%ovalbumin. The grids were then washed four times for 5 min eachtime in PBS-1.0% ovalbumin. For the polyclonal primary antibodies,the grids were exposed for 30 min to 5- or 10-nm protein A gold(Sigma) diluted in PBS-1.0% ovalbumin. Monoclonal primary antibodieswere detected by exposure for 1 h to anti-mouse immunoglobulinG (IgG) (whole molecule) conjugated to 10-nm colloidal gold(Sigma, G-7652) diluted in PBS-1.0% ovalbumin. Double-labelingexperiments were performed by washing the grids four times for5 min each time in PBS-1.0% ovalbumin after the primary antibodyand gold secondary treatment and then treating with the secondantibody and the appropriate gold conjugate as in a single-labelingexperiment. After all labeling steps, the grids were washedfour times for 5 min each time in PBS and three times for 1min each time in distilled water. The virus was then negativelystained with Nanovan (catalogue number 2011; Nanoprobes, Inc.)and air-dried. (Nanovan was used because it produces an intermediatecontrast between glucose and heavy-metal stains. Using glucosewould provide a good contrast with gold and poor contrast withthe particle. Heavy-metal stains [e.g., uranyl acetate] wouldoutline the particle well while almost matching the densityof the gold.) The labeled samples were viewed on a 200-kV fieldemission gun transmission electron microscope (CM series; FEI,Eindhoven, The Netherlands) at a magnification of x38,000 andphotographed on S0163 film (Kodak). Images are represented herein reverse contrast after scanning with a UMAX3000 scanner ata pixel size of 8.33 µm.

Antibody labeling controls. Serial dilution tests were performed with each antibody to determineits optimal concentration for use under our conditions. Littleor no labeling was seen when the antibody was too dilute. Toohigh a concentration of antibody resulted in a high backgroundof gold on the carbon film. All primary antibodies were exposedto protein A gold, protein G gold, or anti-mouse IgG (wholemolecule) conjugated to 10-nm colloidal gold solutions (allSigma products) to ensure their proper secondary detection.In some experiments, mouse mammary tumor virus (MMTV) was mixedwith the PRD1 sample as an internal experimental control. Noneof the antibodies tested labeled the MMTV significantly at theconcentrations that produced labeling on the PRD1.

Omission of the primary antibodies resulted in no labeling ofthe samples or the background. PRD1 mutants lacking specificproteins sus539 (P2^-; 23) and sus690 (P5^- and P2^-; 6) were testedto ensure that the appropriate proteins were absent (Table 1).Two antibodies with presumed internal epitope locations (P7/P14and P16) were also tested on wt PRD1. Neither antibody resultedin significant labeling of the virus.

View this table:
[in this window]
[in a new window]
TABLE 1. Qualitative summary of results for all antibodies and mutant strains tested Immunolabeling results^a

By using a x10 magnifier, gold particles were counted on thenegatives. Areas in which virus was clumped were disregarded,as were background gold particles that were not associated withvirions. Statistical analysis of the labeling data (includingt tests, median calculation, and histogram representation) wasperformed with Microsoft Excel for Macintosh, version 1.0.

Cryo-transmission electron microscopy. All of the virus samples (wt PRD1, sus539, and sus690) wereprepared and imaged under standard holey carbon cryo-grid conditionswith a 200-kV field emission gun transmission electron microscope(CM Series, FEI, Eindhoven, Netherlands) (13, 14).

RESULTS

Top

Results

Microscopy of the preparations. Cryo-electron microscopy (14) showed that most virions wereintact and that the wt PRD1 and sus539 were morphologicallysimilar. Few virions in these preparations showed a single membranoustube associated with the ejection of the DNA. Many virions inthe Sus690 preparation displayed a single tube indicative ofDNA loss.

The cryo-electron micrographs and the negatively stained imagesof the labeled samples showed a consistent morphology for thevirus. The Nanovan-negative stain often penetrated the membraneso that internal features were revealed (Fig. 1). All of thevirus preparations contained small, particulate material thatwas not associated with the virus. A gold marker was often attachedto this particulate material in the images of the anti-P3-labeledsamples (Fig. 1). We believe that this reflects the fact thatour purification procedure did not preserve the intactness ofthe particle perfectly, although it isolated particles efficiently.In addition, particles may have deteriorated after isolationand shed proteins. This material was present on the grids inthe labeling experiments and was what often contributed to the"background labeling" observed in these images. This backgroundlabeling was only observed with anti-P3. The antibody concentrationselected for the other antigens (P2, P5, P6, P11, P20, and P31)resulted in a consistent background that was lower than thecorresponding specific label.

Several controls showed the specificity of the labeling. Omittingthe primary antibody always resulted in a clean background.MMTV was mixed with the PRD1 to act as an internal control (Fig.1B to D, 2A, and 3A) for the specificity of labeling. None ofthe antibodies used showed any significant labeling of the MMTVparticles. Labeling was restricted to PRD1 particles (Fig. 1,2, and 3).

View larger version (97K):
[in this window]
[in a new window]
FIG. 2. Labeling of the sus539 mutant. The sus539 mutant of PRD1 that expresses much reduced levels of P2 (23) was labeled with antibodies against protein P5 (A) and antibodies against P2 (B) that were visualized with protein A coupled to 10-nm gold. All particles of the Sus539 mutant labeled with antibody against P5, while the antiP2 label was concentrated on a small fraction ( $~$ 5%) of the particles.

View larger version (61K):
[in this window]
[in a new window]
FIG.3. Labeling of wt PRD1 with anti-P6 and anti-P20. (A) Mixture of wt PRD1 and MMTV that labeled with antibodies against P6 that was visualized with anti-mouse IgG coupled to 10-nm gold. No labeling is seen on the large, irregular MMTV particles (arrow), and a single site of labeling is seen on individual PRD1 particles. (B) Labeling of wt PRD1 with anti-P20 visualized with protein A coupled to 5- nm gold. (C) A double-labeling experiment in which wt PRD1 was labeled with antibody to P20 and visualized with protein A coupled to 5-nm gold and then with antibody to P6 visualized with 10-nm gold coupled to anti-mouse IgG.

The labeling results are summarized qualitatively in Table 1,and representative images are shown in Fig. 2, 3, 4, and 5.The observation that different antibodies labeled the virionin distinct ways (i.e., anti-P3 labeled the whole of the virionsurface, anti-P5 labeling was consistent with its labeling ofmost vertices, anti-P6 labeled only a single vertex, and nolabeling was observed with antibody to P16; Table 1) was a furtherindicator of the specificity of our immunolabeling procedures.All specific labeling values exceeded the density of gold attachmentto the film between particles and the density of the nonspecificlabeling of MMTV.

View larger version (194K):
[in this window]
[in a new window]
FIG. 4. Double labeling of PRD1. (A) wt PRD1 was labeled with antibodies against P5 visualized with protein A coupled to 5-nm gold and then washed and labeled with antibody to P2 that was visualized with protein A coupled to 10-nm colloidal gold. Panel B shows the corresponding experiment in which wt PRD1 was labeled with antibodies to P5 and visualized with protein A coupled to 5-nm gold, washed, and then labeled with antibodies to P6 and visualized with 10-nm gold coupled to anti-mouse IgG. (C and D) Labeling of wt PRD1 (C) and the Sus539 mutant (D) with antibody to P2 visualized with protein A coupled to 5-nm gold and then with antibody to P6 visualized with 10 nm of gold coupled to anti-mouse IgG.

The sus539 (P2-deficient) mutant gave an unexpected result whenlabeled with an anti-P2 antibody (Table 1). Most of the virionshad no label at all, as expected. However, a few of the virionshad a labeling pattern similar to that of wt PRD1 labeled withanti-P2, with gold markers at most vertices (Fig. 2). As describedbelow, we believe this reflects leakage of the mutant in a fractionof the cells.

Single labeling showed a marked difference between the antibodiesthat labeled multiple vertices, such as those against proteinsP5 (Fig. 1C), P2 (Fig. 1D), and P31 (Fig. 5A), and the antibodiesthat labeled a single vertex, such as those against proteinsP6 (Fig. 3A), P20 (Fig. 3B), and P11 (Fig. 5B).

View larger version (104K):
[in this window]
[in a new window]
FIG. 5. The binding and translocation vertices. The labeling of wt PRD1 with antibodies directed against a binding vertex component (A) (anti-P31) contrasted with labeling of translocation vertex components (B) (anti-P11). The anti-P31 in panel A was visualized with protein A coupled to 5-nm gold. P31 is known to be involved in the receptor binding complex. The anti-P11 in panel B was visualized with anti-mouse IgG coupled to 10-nm gold. P11 is known to be involved in the DNA delivery that leads to infection and translocation of the genome into the target cell.

We used double labeling (Fig. 4) to test whether the single-labelingresults reflected the alterations in labeling efficiency ora variation in protein localization. The availability of antibodiesfrom different species allowed us to visualize separate antigenswith differently sized gold labels. We believe that the factthat we did not see multiple labels at a single vertex indicatesthat there was not enough room either for two primary Igs orfor two gold secondary markers (Fig. 4). Single-labeling experimentswith these antibodies resulted in only one gold marker per vertex(Fig. 1C and D, 2, 3A and B, and 5). Steric hindrance can preventthe occurrence of two gold markers at a single vertex but, ofcourse, cannot prevent different labels binding to differentvertices.

Double labeling of wt PRD1 with antibodies to P2 or P5 thatlabeled multiple vertices when applied singly did not preventthe subsequent labeling with antibodies to P6. This indicatesthat P6 is at a different vertex from that labeled with P5 andP2 (Fig. 4). In contrast, we did not observe labeling with antibodiesto P6 and P20 in the same wt PRD1 virion (Fig. 3C) under double-labelingconditions. If the P20 and P6 were at different vertices, wewould have expected to see two gold labels (at different vertices)per virion with the frequency expected from single labeling.Labeling with antibodies to P20 appeared to block subsequentlabeling with antibodies to P6. We concluded that P20 and P6must be located at the same vertex.

Quantitation of the labeling results. Our labeling conditions did not result in quantitative labelingof all of the vertices. Nevertheless, quantitation of the labelingresults (Fig. 6) reinforced the impression given by the individualimages. It also allowed us to define the limits for the reliabilityof our results and to address the effect of differing efficienciesof labeling on our results. The variation in labeling appearsless dramatic when one considers that the orientation of thevirion on the grid can shield a potential binding site and thatthis effect is more pronounced in the case of the unique vertexantigens.

View larger version (30K):
[in this window]
[in a new window]
FIG. 6. Quantitation of labeling results. The percentages of virions are plotted against the number of sites labeled on each particle. Panel A shows the distribution for antibodies that label the initial binding vertices (anti-P2, anti-P5, and anti-P31). Panel B shows the distribution for the antibodies that label the translocation vertex (anti-P6, anti-P11, and anti-P20).

Figure 6A documents the labeling of multiple sites on individualparticles by antibodies against P31, P5, and P2. Antibodiesto P31 produced an average of 4.6 gold particles (median, 4.5;standard deviation, 1.7) per virion. Antibodies to P5 producedsimilar labeling (mean of 5.7 gold particles; median, 5; standarddeviation, 2.0) per virion. Both of these antibodies labeledall of the particles. The antibody against P2 produced somewhatlower labeling, with an average of 2.4 gold particles (median,2; standard deviation, 1.0) per virion. Only 94.6% of the particlesshowed labeling with anti-P2. No more than 10 sites were labeledwith any of the antibodies on any particle. This presumablyreflects the lack of exposure of all vertices when the particlelies on the grid.

Antibodies against P6, P20, and P11 showed a completely differentlabeling pattern. The extent of labeling was much lower withthese antibodies. None of the virions were labeled at multiplesites (Fig. 6B). Antibodies against P6 labeled 19.2% of theparticles (n = 297). Less labeling was seen with antibodiesagainst P20 (7.8%; n = 488) and P11 (10.3%; n = 341). If labelingof a second site were independent of the labeling at the first,one would expect to have seen two or more labels on a fractionof the particles in the case of P6 (8.8 double labels expectedin 297 particles) since the frequency of single labeling isrelatively high. The lower frequency of labeling with antibodiesagainst P11 and P20 would still have allowed the detection ofmultiple labeling if it were present (3.2/341 expected for P11and 2.4/488 expected for P20) and the labeling of the siteswere independent. We conclude that only a single site was labeledin the particle by any of these antibodies.

The result with the sus539 mutant was unexpected. Since thismutant should be deficient in P2, we expected no labeling. Wesaw labeling in 4.5% of the particles (n = 852); however, thedistribution of the gold found on this labeled fraction showedcharacteristics similar to those of the wt (mean = 3.0; median= 2; standard deviation = 2.0 gold labels per particle). Oneexplanation of this effect is that the amber mutation responsiblefor the lack of synthesis of P2 in the mutant cells is suppressedin a small fraction of the cells, which then give rise to wtvirus. In addition, amber mutants such as sus539 are known tobe leaky. The ion exchange chromatography may enhance this phenomenonby selecting the P2-containing virions always present in thevirus stocks. This observation gives insight into the mechanismof expression, since it suggests that different cells in thepopulation react differently. This phenomenon could only beobserved with a method that can examine the composition of individualparticles.

The use of double labeling indicates that antibodies to P5 andP2 label independently of P6. Antibody to P5 labeled the virionat 4.8 ± 0.4 sites in the double-labeling experimentand shows an efficiency similar to that obtained in the single-labelingexperiment. Antibody to P6 labeled 23% of the virions in thedouble-labeling experiment, giving efficiency similar to thatshown in single labeling. Antibody to P2 showed efficiency similarto that seen in the single-label experiment (P2, 2.4 labels/virion;P6 labels 17% of the virions). A very different picture wasseen when double labeling was performed with antibodies to P20and P6. P6 and P20 labeling were decreased in the double-labelsituation to 13.3 and 9.6% of the virions, respectively. Thesum of P20 and P6 labeling (22.9%) approached that of singlelabeling with antibodies to P6. We concluded that antibodiesagainst P5 and P2 label at an independent site from that labeledby antibodies to P20 and P6 and that labeling with antibodiesto P20 interferes with labeling of P6.

DISCUSSION

Top

Discussion

Our labeling results with PRD1 are simply summarized. An antibodyto P3 labeled many sites on the virion surface. This is consistentwith the biochemical and structural results that show it tobe the major structural protein forming the virion shell (8,13, 14, 46). The labeling with antibodies directed against proteinsP2, P5, and P31 was consistent with their presence at most ofthe vertices. The labeling with antibodies to P6, P11, and P20supported their localization to the same unique vertex. Doublelabeling showed the unique nature of this vertex in that labelingwith antibodies to P20 inhibited subsequent labeling with P6.Labeling with antibodies to P2, P5, or P31 did not inhibit labelingwith antibodies to the unique vertex component P6.

We are able to interpret our labeling results functionally becauseof the wealth of biochemical and genetic data that are availablefor this membrane virus. Our data and recent genetic evidence(53a) support a division of the 12 vertices into two classes.Of the 12 vertices, 11 correspond to the location of the threeproteins in the receptor binding complex. We refer to this classas the initial binding vertex. The later events in infectionthat result in penetration of the DNA through the target cellwall occur through the twelfth vertex that is also responsiblefor the initial packaging of the genome. We refer to this asthe translocation vertex. Whether any of the unique vertex proteinsare involved in the initiation of particle assembly to assurethe presence of the single vertex is an intriguing question.P11- and P20-deficient mutants are not compromised for particleassembly. This question will remain open until a P6-deficientmutant becomes available.

How does this arrangement of initial binding and translocationvertices function? This work, in combination with recent geneticevidence (53a), shows that P6, P9, P20, and P22 are all involvedin packaging and reside at the same vertex. The presence ofP11 indicates that it is also the site of DNA exit (24). Itmay seem paradoxical that the major receptor binding protein,P2, is found at the other 11 vertices of the particle; however,it reveals the nature of the interaction of this membrane viruswith the host.

The fact that the initial binding is reversible (23) and thata further transition must occur prior to permanent binding andinfection is consistent with the infection pathways of othermembrane-containing viruses such as influenza virus, SemlikiForest virus, and human immunodeficiency virus (19, 30, 48,49). In these mammalian viruses, the process of infection ismultistep. A weak initial binding concentrates the virus nearthe target cell and allows it to search the cell surface fora productive binding site (36, 38, 53a). Only at the correctand productive binding site does the virus trigger its permanentbinding and infection. The distinct multiple vertices of PRD1accomplish searching followed by triggering similar to thatseen in other membrane viruses.

ACKNOWLEDGMENTS

We thank Barbara Watson (Structural Biology Division, WellcomeTrust Centre for Human Genetics, University of Oxford) for themouse mammary tumor virus stock. We thank our colleagues JohnBriggs and Jean-François Trempe (Wellcome Trust Centrefor Human Genetics) for useful discussions and a critical readingof the manuscript.

We are pleased to acknowledge the support of the Wellcome Trust(HH5FE) for this work. S.D.F. is a Wellcome Trust PrincipalResearch Fellow. This investigation was supported by researchgrant 172904 (Academy of Finland; J.K.H.B.) and research grants168694, 164298, and 172621 (Finnish Centre of Excellence Programme,2002-2005; D.H.B.) from the Academy of Finland. The collaborationbetween our laboratories was supported by a Human FrontiersScience Programme Grant (RGP0320/2001-M).

FOOTNOTES

* Corresponding author. Mailing address: The Division of Structural Biology, Wellcome Trust Centre for Human Genetics, The Henry Wellcome Building for Genomic Medicine, University of Oxford, Roosevelt Dr., Headington, Oxford OX3 7BN, United Kingdom. Phone: 44-0-1865-287546. Fax: 44-0-1865-287547. E-mail: stephen.fuller{at}strubi.ox.ac.uk.

REFERENCES

Top