INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In the preceding paper (10a), we outlined the theoretical and experimental arguments in support of our cisternal-wrapping model of vaccinia virus assembly (36). The data provided in the preceding paper also show that during assembly the membranes of the intracellular mature vaccinia virus (vaccinia IMV) and the underlying core are folded upon each other via a process reminiscent of gift wrapping, but instead of having a planar organization, the virus appears to be made up of intricate tubular-cisternal domains that together make a complex labyrinth.

In the present study, we used thin-section electron microscopy (EM) in combination with immunocytochemical labeling to focus in more detail on the stages of vaccinia virus assembly, from the initial small crescent to the final IMV. A background to the key historical references detailing analyses of vaccinia virus assembly using thin-sectioned EM material is given in the introduction of the preceding paper (10a).

Here, we emphasize the following points about the assembly of vaccinia IMV. First, the endoplasmic reticulum (ER)-derived cisterna that will assemble into the virus has a propensity to collapse tightly upon itself, giving the impression of a single bilayer in cross-sections of this flattened cisternae. Second, the tubules that we showed in the preceding study to be an intimate part of the IMV structure can be detected in continuity with the crescent at very early stages of assembly; these 30- to 40-nm-diameter tubules have been extensively described in the literature on vaccinia virus assembly (3, 12, 20, 28, 36). Third, we extend our earlier model in which the viral DNA preassembles on smooth-ER membranes prior to entering the assembling particle. These membranes are continuous with the viral envelope in what we argue is one cisternal structure that folds upon itself to assemble the IMV.

The use, and especially the interpretation, of thin sections for EM is not a subject to be taken lightly, especially with structures with the complexity and the relatively small size of vaccinia virus. When a thin section is produced, the loss of (most of) one dimension occurs; this often makes profiles difficult to interpret (5, 7, 38). The physics of cutting sections has been widely discussed but is poorly understood (see references 4 and 10). The material is always significantly compressed, at least transiently, during this process, although in some cases plastic reformation may occur completely. Interpretations of which molecules bind heavy-metal stains, such as osmium or uranyl acetate, are always questionable when one is dealing with membranes, since no one really understands what molecules in the bilayer are responsible for the unit-membrane appearance. For example, even quantitative removal of membrane lipids of chloroplasts and mitochondria still gives clear unit membranes with osmium tetroxide staining and epon thin sections (see reference 7 for a discussion of this point). These kinds of observations are at odds with the claim by Hollinshead et al. (12) (in support of a one-membrane model for vaccinia IMV) that the usual trilaminar appearance of membranes in plastic sections is invariably due to binding of heavy-metal ions to phospholipid headgroups. It should be noted that the molecular details of conventional staining methods for EM were the subject of vigorous debates in the 1950s ands 1960s. Although these debates are now forgotten, the relevant issues have remained unresolved (see reference 7).

Despite these complications, if one wants to see the details and label antigens in the central parts of the virus, one is obliged to use thin sections, obtained either by traditional plastic-embedding methods or with thawed cryosections. Both of these approaches depend on the use of aldehyde fixatives, which have the potential to significantly alter membrane structures. We feel confident that this is unlikely to be a significant problem for vaccinia IMV or its assembly intermediates for two reasons. First, a recent study of vaccinia virus-infected cells prepared by freeze-substitution (in which the cells are rapidly frozen and fixed at low temperatures) revealed no significant differences in the various assembly intermediates or in the IMV versus conventionally fixed cells (27). Second, our extensive cryo-EM studies (10a, 29) show that the structure of the IMV prepared without fixatives is not significantly different from that of virus prepared with fixation.

Since the question of what structure is, or is not, a membrane is absolutely crucial to the question of IMV structure and assembly intermediates in the present and the preceeding papers, we again use specific antibodies against the cytoplasmic domains of well-characterized viral membrane proteins. Since these proteins are very abundant, the pattern of labeling allows one to follow the pattern of membrane profiles more convincingly. When these antigens are enriched in membrane tubules, they are exposed on the outer (cytoplasmic) surface of the tubule, where the labeling is especially prominent. We emphasize that this labeling approach cannot unequivocally help us to distinguish a single bilayer from a tightly apposed cisterna. As explained in the accompanying paper (10a), all of the existing evidence favors a cisternal model and the idea of a single membrane does not fit into any known cell biology paradigm. We start the second paper by providing further support for this crucial point.

To label the viral membranes, we again took advantage of two well-characterized IMV membrane proteins, P21 (A17L) and P16 (A14L), and for one micrograph P8 (A13L), for which well-characterized polyclonal antibodies are available. We argue that strong labeling of a membrane-like structure with both antibodies provides incontrovertible proof of the existence of a virally modified membrane or membranes. Identification of viral membranes is particularly important in the case of viral membrane tubules that are continuous both with the rough ER (RER) and with the crescent membranes (18, 31, 36) (see below). The dimensions of most of these tubules are such that when tubules are clearly identified in thin sections, they are almost certainly embedded within the sections; under these conditions, luminal antigens would be totally inaccessible. Since all of these antibodies recognize the cytoplasmic domains of their respective antigens, they can be expected to label the outside of the viral tubules (see also references 18 and 31) An additional problem with such tubules is that connectivities to other viral membranes are always seen in projection rather than as direct continuities.

Most scientists would probably suggest that the easiest way to understand the structure of a virus would be to serial section and "simply" reconstruct the structure (as implied by Hollinshead ands colleagues [12]). If one could produce almost infinitely thin sections through vaccinia virus, this might be a reasonable approach. However, it must be realized that although vaccinia virus is a relatively large object, as far as viruses go, there are practical limitations. In two to six thin sections (50 to 60 nm), one cuts through the entire virus, depending on the orientation. In addition to compression, further difficulties, such as the possible flow of supporting resin and the unsolved physics of the sectioning process, make it very difficult to accurately mount profiles on top of one another during reconstruction attempts (although improved technology is now available [see below]). These were especially evident in the first serial-section analysis of vaccinia virus, performed by Morgan et al. (21), which provided little three-dimensional information, and none that was not already obvious from thin-section analysis. Our attempts to interpret IMV structure by using serial plastic sections also led us nowhere. We also strongly dispute the claim of Hollinshead et al. (12) that the question of whether the outer layer of the IMV is a single or double bilayer can be resolved simply by tilting the specimen.

The recent introduction of new technology combining cryofixation and dual-axis tomography offers many potentially important new approaches to overcome the traditional problems involved in three-dimensional reconstructions of structures such as vaccinia virus (2, 14, 19). Studies using such an approach are now in progress. Without such technology, however, it is impossible to predict how far one can interpret single- and double-membraned profiles in thin sections via EM.

The previous paper (10a) focused on the structure of the isolated IMV and on its disassembly during entry or following treatment with dithiothreitol (DTT). Here our aim was to complement those three-dimensional data by examining thin slices through the IMV and, in more detail, through the intermediate stages that start with the viral crescents.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cells, virus and antibodies. The HeLa cells and their infections with vaccinia virus were described by Krijnse Locker et al. (15, 17). The antibodies against P16 (A14L), P21 (A17L), and P8 (A13L) have been described previously (23, 31). The permeabilization of infected cells with streptolysin O (SLO) was described by Krijnse Locker et al. (16). Briefly, at 6 to 8 h after infection, cells were rinsed with ice-cold phosphate-buffered saline and once with SLO buffer (25 mM HEPES buffer [pH 7.4], 115 mM potassium acetate, 2.5 mM MgCl₂) containing 1 mM DTT. SLO (purchased from S. Bhakdi, University of Mainz) was added to this cold buffer at 1 µg/ml and left on the cells for 10 min. Subsequently, the cells were rinsed with SLO buffer containing DTT but lacking SLO and warmed to 37°C for 10 to 15 min to extract cytoplasmic proteins. The cells were then fixed with 1% glutaraldehyde and processed for EM.

EM. Fixation and preparation of cells for epoxy resin embedding and for cryosections has been described previously (10a, 36). The preembedding labeling for p8 (A13L [31]) was done as described by Krijnse Locker et al. (18). For a general overview of these techniques, see reference 7.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Thin sections: conventional views. As shown in the accompanying paper (10a), the structure of the IMV is incredibly complex. To facilitate our description of thin-sectioned profiles, we start by presenting micrographs from conventional epoxy resin sections, as well as a Tokuyasu cryosection, in which the IMV membranes are always more distinct. These images, from HeLa cells infected for 8 h, serve as a reference point.

View larger version (177K): [in this window] [in a new window]   FIG. 1.   Illustration of the assembly problem as seen in thin sections of vaccinia virus-infected HeLa cells at 8 h postinfection. (A, inset) Four consecutive stages in assembly of vaccinia virus (from the EM negative of an epoxy resin section). From left to right, a spherical immature virus (IV) (1) condenses to form a more electron-dense, still-spherical intermediate (2). The latter rearranges into an IMV (3), whose projection in this image is oval. The IMV becomes engulfed by a trans-Golgi network-derived cisterna to form the IEV (4). Arrows indicate the two cisternal wrapping membranes. (A) An epoxy resin section of different stages in the assembly of the crescents (arrowheads) and the IVs, which are usually perfectly spherical in profile. The large arrow indicates the dense nucleoid of DNA within an IV profile. The small arrow indicates the spicules. (B) A relatively thick thawed cryosection of an infected cell showing different profiles through the IMV. One particle (star) shows a classical brick-shaped profile whose two membranes are indicated by arrowheads. Note that all other profiles through IMV particles show different shapes, depending on the plane of sectioning. However, in most cases the two distinct membrane profiles (inner and outer) are easily identified. The small arrow indicates the spicules. Bars, 100 nm.

Sections of SLO-permeabilized cells and labeling of viral membrane proteins. Since the presence of cytoplasmic proteins often obscures fine details in thin sections of cells, for many experiments we took advantage of the use of SLO to selectively permeabilize the HeLa cells and thereby remove cytoplasmic proteins by incubating the permeabilized cells for a few minutes in buffer before fixation (16). This procedure greatly increases the contrast of membranes and other structures; it can also increase accessibility of antibodies to antigens that are on the cytoplasmic surface of membranes when thin sections of such cells are labeled. The fact that many complex in vitro processes can be reconstituted in such permeabilized cells indicates that this procedure is not likely to destroy cellular structure.

View larger version (125K): [in this window] [in a new window]   FIG. 2.   Details of early crescent formation at 6 h postinfection in HeLa cells that were permeabilized with SLO prior to fixation. (A and B) Epoxy resin sections showing connectivities (large arrowheads) between viral crescents (small arrowheads) and cellular ER; the ribosomes are indicated by arrows. Note that the crescent cisternae collapse to give the appearance of only one membrane existing in the crescent domain. (C and D) Thawed cryosections through two adjacent and oppositely oriented (as a result of the infolding process that leads to IMV assembly) crescent domains (arrowheads). The sections in panels D and E were labeled with anti-P16 (cytoplasmic domain). The arrows in panel D indicate membranes in continuity with the viral crescent (arrowheads). Note that the label is well separated from the crescent membrane in both panels D and E and that in panel E it extends into the tubular extension (arrows) that emanate from the inner, concave aspect of the crescent. Bars, 100 nm.

View larger version (137K): [in this window] [in a new window]   FIG. 3.   Epoxy resin sections of crescents from SLO-permeabilized infected cells at 6 h postinfection. In these images, the crescent domains (large arrowheads) are in continuity with cisternal (arrows) and tubular (small arrowheads) domains. The individual membrane bilayers can be clearly seen in these images (arrows). Note also the propensity of the crescent to form an S-shaped structure (Fig. 2C), as well as the apparent requirement of two distinct crescent domains for assembly into the precursor of one IV (C and E). Bars, 100 nm.

View larger version (114K): [in this window] [in a new window]   FIG. 4.   Visualization of viral membranes in thawed cryosections of vaccinia virus-infected cells at 6 h postinfection, using anti-P21 as well as anti-P16 antibodies. (A) A low-magnification overview of two IVs single labeled for P16. Note the extensive tubular membrane labeling that connects the ER network with developing IVs. (B to E) sections were double labeled with anti-P16 (10-nm gold particles) as well as anti-P21 (5-nm gold particles). Note the uninterrupted progression of label from the crescents and IVs to the connecting membrane tubules. (E) IV has enclosed DNA, while a second domain of DNA (asterisk), with its 2.5-nm periodicity evident, is seen adjacent to the IV, in close to P16-labeled membranes. Bars, 100 nm.

View larger version (158K): [in this window] [in a new window]   FIG. 5.   (A) An Epon section of infected cell (6 h postinfection) that had been labeled with anti-P8 and 5-nm gold particles before being embedded (after permeabilization). Note the S-shaped labeled cisterna whose membrane is mostly distinct. However, at a few sites (arrowhead), the two cisternal membranes show a local tendency to collapse (arrowhead). (B to E) Epoxy resin sections showing the close relationship between the viral DNA (which has the characteristic 2.5-nm periodic repeat and is very electron dense), the viral crescents, and the cellular ER. (B) DNA (D) is in the dense factory region, adjacent to a forming IV (the arrowhead indicates continuity between the crescent domain and a tubular element [arrow]). (C) At 10 h postinfection, there is extensive ER that is closely apposed on much of the surface of the DNA. The arrowhead indicates close continuity or contiguity between the ER next to the DNA (D) and a viral crescent. Note also that the profiles of DNA in panel B and in some parts of panel C are not in obvious contact with the ER. Details of the close apposition of ER cisternae with viral DNA are evident in panels D and E (8 h postinfection). Bars, 100 nm.

DNA entry intermediates and sealing of the IMV. The remaining epoxy section micrographs, in Fig. 5B to E, document the appearance of the viral DNA and its relationship to the crescents, as well as to the ER. Figure 5B shows the DNA apparently free of membranes, adjacent to an IV that also shows the inward folding of a tubule emanating from the crescent. In Fig. 5C, different pieces of DNA apparently free of membrane are seen (right side of the image), bounded on one or both sides with ER membrane (center of image). There is evidence of a likely continuity between a crescent and a part of the ER that is attached to the DNA. In Fig. 5D and E, discrete DNA structures are sandwiched between closely apposed ER cisternae. It should be noted that the non-membrane-bound parts of these DNA structures end rather abruptly; we presume that at these sites the DNA may end up in loops.

View larger version (111K): [in this window] [in a new window]   FIG. 6.   Details of close apposition of crescents or IVs, DNA, and P16- and P21-labeled membranes. In all images, DNA is indicated by D, P16 is labeled with 5-nm gold particles, and P21 is labeled with 10-nm gold particles. The low degree of labeling of the crescent in panel C is probably due to the fact that the periphery of the crescent is embedded within the section and therefore inaccessible to antibody. Note that the crescent in panel F (large arrowhead) shows two distinct membrane profiles. The small arrowhead in this figure indicates a cross-section of a tubular membrane profile. In panels D and F, the DNA profiles are not intimately associated with P16- and P21-enriched ER domains, while in panels B and E these domains are seen in close contact with parts of the DNA. In panel F, the two membrane profiles of a beginning crescent are distinct and are indicated by two arrows. Bars, 100 nm.

View larger version (198K): [in this window] [in a new window]   FIG. 7.   (A) Cryosection (8 h postinfection) of an infected cell that was double labeled for P21 (10-nm gold particles) and P16 (5-nm gold). Note the labeling of membrane profiles closely attached to the DNA and emanating from this region. The ER domains directly adjacent to the DNA have only two gold particles for P21 and one for P16; these proteins are evidently mostly excluded from these domains. (B and C) Plastic sections showing different views of the viral DNA entering the IV. (B) Membrane profiles are not evident near the DNA (arrowhead). (C) The electron-dense DNA is closely attached to membrane tubular structures that seem to enter the IV (arrow). The large arrowheads on the right of this image show putative DNA elements adjacent to a developing crescent (small arrowhead). Bars, 100 nm.

View larger version (148K): [in this window] [in a new window]   FIG. 8.   A gallery of different intermediates in the process of DNA entry into the IV, and some aspects of the complex morphogenetic changes that accompany the conversion of the spherical IV into the more-brick-shaped IMV. (A, G, H, J, and K) Epon sections; (B to F and I) cryo-sections. Note that the DNA profile (D) inside the IV in panel A is not surrounded by visible membranes, whereas that in panel B it is lined on two sides by membrane profiles (arrowheads). The latter develop into the membrane profiles that eventually surround the core (arrowheads in panels C to J). The arrows in panel J indicate possible tubular extensions of a stage just before the assembly of the IMV. In panel K, the particle appears almost like an IMV except that a large tubular or cisternal extension (arrow) has not yet attached closely to the particle. The gold in panel C shows labeling for P16; in this image, the separation of the membranes is sufficient to conclude that two of the gold particles are associated with the developing core membrane. Bars, 100 nm.

Labeling of profiles of IMV in infected cells. To focus in more detail on the appearance of the IMV in section profiles, we performed double immunolabeling for P16 and P21 on thawed cryosections of vaccinia virus-infected cells (8 h postinfection) (Fig. 9). This image can be compared with the unlabeled image shown in Fig. 1B. Considering first only structural aspects, these images show, again, how the appearance of the particle varies greatly in different, random sections, although the outer cisterna and the core cisterna are clearly evident in most particles. In sections parallel to the "top" of the particle (Fig. 9A), the two cisternal membranes are almost parallel around the particle. The center of this particle shows a membrane cisterna that has been "shaved," exposing a stain-excluding membrane structure which represents the top of the core. In orthogonal sections, the particle reveals an oval to dumbell-shaped core profile that has been widely described (Fig. 9 A, inset; Fig. 9B). In some sections are seen one to three spherical profiles (Fig. 9 A, inset; Fig. 9B) that correspond to highly curved tubules described by Peters and Mueller (26) and in the accompanying paper (10a).

View larger version (140K): [in this window] [in a new window]   FIG. 9.   Cryosections of IMV in an infected cell (8 h postinfection) that have been double labeled for P16 (10-nm gold particles) and P21 (5-nm gold particles). (A) One particle has been sectioned parallel to and just beneath the "top" of the particle (as defined in the accompanying paper [10a]); the star indicates the projection of a viral core lobe that is embedded within the section. The outer membrane profile and the core membrane profile are indicated by large arrowheads. The small arrowhead shows an oblique section revealing the continuities between these curved tubular membrane structures and more extended regions. In the inset, the ~50-nm circular profiles in the core structure, seen in favorable sections, are indicated by arrows. In panel B, one profile shows three of these spherical structures (arrows). It is difficult to precisely determine the localization of gold particles to particular membrane profiles in these images because of overprojection problems (the gold labeling is predominantly on the surface of the section, whereas significant amounts of viral structures are embedded within the bilayer). These images strongly suggest that the labeling of both P16 and P21 extends to the core membrane. Bars, 100 nm.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The thin-section analysis we have presented in this and in preceding papers fit well into our overall scheme (see Fig. 2 in the accompanying paper [10a]). The data collectively argue that vaccinia virus assembly is initiated by the assembly of new domains in specialized smooth-membrane regions of the cellular ER network. We suggest that this is a consequence of lateral segregation of the viral membrane proteins in these domains, which have the innate capacity to segregate the cisternal and tubular subdomains, all in continuity with the cellular ER. All existing evidence suggests that this segregation excludes all cellular membrane (and nonmembrane) proteins (22), thus fitting well into a general strategy used by many membrane viruses, enrichment of viral membrane proteins and exclusion of host proteins during assembly (9, 33).

A surprising conclusion of our model is that in essence the IMV topologically consists of a single membrane in continuity with itself. While the title of the Hollinshead et al. (12) paper is thus correct, the details are quite different. Whereas for the latter authors, and in Dales' view, the DNA is enclosed within this vesicle, in our view the flattened vesicle (two membranes on each other) has the DNA on its outer surface. In our model, both in the interface with the DNA and on the outside of the IMV it is the cytoplasmic leaflet of the bilayer that is exposed to the outside world. This is a unique feature of poxviruses and related viruses compared to membraned viruses in general. Other membrane viruses, as well as the extracellular enveloped vaccinia virus (EEV), invariably expose their luminal domains on their surfaces.

Vaccinia virus and its distant relative African swine fever virus (ASFV) have many similarities in terms of their modes of assembly. Most importantly for our vaccinia virus assembly model, ASFV begins its assembly by a process that unequivocally involves ER wrapping (1, 30). For some reason, the cisternal membranes in this virus do not flatten as do those of vaccinia virus. This phenomenon makes the ER wrapping data indisputable in the case of ASFV and, by analogy, provides compelling support for this model for vaccinia virus. Given their many similarities in a number of complex processes, ASFV and vaccinia virus must surely have shared a common ancestor in the distant past.

We have provided many lines of data that support our model that the crescent and IV "single" membrane profile represents a flattened cisterna which, at least early during assembly, is functionally and structurally continuous with smooth-ER membranes; the latter are themselves continuous with the intermediate compartment (IC) between the ER and the Golgi complex (6, 18, 31, 36). (At that time we referred to the crescent domain as being connected to the intermediate compartment [IC], which was then a new and relatively poorly characterized structure. While there is still dispute about whether this organelle is a separate entity or is a subdomain of the ER (our view) or of the Golgi [see reference 8], it is generally agreed that the IC is the last station for export from the ER to the Golgi apparatus. We therefore prefer now to use the term smooth ER for the specialized membrane domain which develops into viral membranes, to avoid giving the (false) impression that this modified viral domain functions in the export pathway from the ER to the Golgi complex.) In agreement with this ER-derived cisternal model, when two vaccinia IMV membrane proteins (but not the six EEV proteins known from other studies) were expressed by themselves in uninfected cells, they were shown to be efficiently targeted to pre-Golgi, smooth-ER membrane subdomains (reference 18 and unpublished data). We have also provided evidence that the IMV cisterna does not fuse with itself to form two continuous layers of membrane around the virus but is rather folded upon itself (29). We suggested that the particle is sealed by a proteinaceous (or Velcro-like) plug that glues one cisterna upon itself (29). One argument for such a model is the fact that trypsin treatment of IMV allows access of the protease, as well as uranyl acetate, into the core interior, a finding first described by Peters (24, 25). If the IMV was completely sealed by a single membrane layer, it could be difficult to imagine how such a treatment could "open up" the particle.

We had previously argued that the viral DNA associates closely with smooth-ER subdomains prior to entering the late IV stage particle (6). Similar data were evident (but not explicitly pointed out) in earlier work (11). This association is especially pronounced after treatment with rifampin (which blocks IV assembly) or in a temperature-sensitive mutant, ts16, at the nonpermissive temperature; under these conditions, assembly is blocked at a late IV stage in which the DNA partially enters the particle (6). All of our evidence argues that the smooth-ER domain that binds the DNA excludes both cellular and, with one exception, all viral markers that we have tested using antibodies. The exception is the membrane protein E8R, which localizes to some of these membrane domains, as well as to the core membrane (L. Doglio, S. Schleich, and J. Krijnse Locker, unpublished data). In addition, in this study we found small but significant amounts of the viral membrane proteins P16 and P21 in these membranes, perhaps an indication of incomplete sorting but, more importantly, providing functional evidence for continuities between these domains and other virally modified smooth membranes. In this paper, we have also provided additional structural evidence for such continuities.

Our model predicts that the virally modified membrane domains that form the crescent, the internal tubular-cisternal membranes of the IV, and the membrane that binds the DNA are distinct but in direct continuity. That ER membranes can segregate into different domains is known from the observations that fundamentally different domains, such as the rough ER, smooth ER, and inner and outer nuclear envelopes, can be structurally in direct continuity, all parts of the greater ER network.

From our topological model (Fig. 2 in the accompanying paper [10a]) we now propose the following sequence of events for the assembly of the IMV. From considerations of the topology of the known proteins, very little protein mass is found in the luminal domain, and there are no luminally glycosylated IMV proteins. This presumably facilitates the cisternal flattening that is evident. The key processes are as follows:

(i) Viral membrane proteins segregate into separate membrane domains of the ER via lateral self-aggregation, and host proteins are excluded.

(ii) At least three different extended domains are assembled. First is the crescent curved cisternae. As suggested here, two distinct crescent domains may be involved for each virus particle assembled; these are connected to each other via S-shaped membrane links (see, e.g., Fig. 3C and E). Second and third, respectively, are the tubular and cisternal domains that are continuous with the inner membranes of the crescent that we argue develop into the tubulo-cisternal domains that form a partial fold around the DNA.

(iii) These different domains, which presumably initially assemble as functionally distinct entities, must then cooperate via a remarkable process of self-organization in which DNA enters a particle that folds upon itself, thereby effectively sealing the particle by a mechanism that is sensitive to exogenously applied proteases, as well as to reducing agents (15, 29).

The convex (outer) side of the membrane crescent cisterna fails to be labeled by any of our antibodies against IMV membrane proteins until it is at a fairly advanced IV stage (35). At this point, at which the electron density of the late IV increases, the outside of the particle rapidly acquires a high concentration of the peripheral membrane protein P14. The interior, concave aspect of the crescent is enriched from the first stages of assembly in the peripheral membrane protein P65, the target of rifampin, while the outer aspect, as well as the extended tubules, are mostly unlabeled. This protein, which we have proposed to function as a temporary scaffold that stabilizes the highly curved crescent, is degraded at about the time when the particle is sealed (13). The other domains we can identify is made up of tubules and loose cisternal domains; these, as well as the inner aspect of the crescent, can be labeled on their (exposed) cytoplasmic surfaces for the membrane proteins P16 and P21 (18, 31; this study). Another membrane protein, P8, is present not only in these membranes but also on the outer surface of the crescent and IMV (31). As shown here (Fig. 5A), this protein associates with curved cisternae before the latter flatten into the crescent structures.

The ability of proteins to be localized to outer and/or inner domains of the crescent and IV can be easily rationalized by our cisternal wrapping model, in which it is possible that membrane proteins diffuse from one side of the cisterna to the other. This same phenomenon was particularly evident with P16, which is usually excluded from the outer membrane of the cisterna; when the viral structure is "loosened" with DTT, this membrane protein can diffuse to the outer layer of the IMV (see accompanying paper [10a]). We have at present no molecular information on how the different viral membrane domains cooperate to drive the assembly process.

In the third study in this series, we take advantage of a vaccinia virus mutant lacking a key abundant core protein, P4a, in which assembly is arrested (or slowed down) at a key step or steps in the DNA entry process. These preparations provide access to a process that normally must happen very quickly, since the intermediates (e.g., those shown in Fig. 7 and 8) are normally quite rare.