Quasi-equivalent viruses: a paradigm for protein assemblies1

J. Mol. Biol. (1997) 269, 665±675

REVIEW ARTICLE

Quasi-equivalent Viruses: A Paradigm for Protein Assemblies John E. Johnson* and Jeffrey A. Speir Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037 USA

The structure and assembly of icosahedral virus capsids composed of one or more gene products and displaying quasi-equivalent subunit associations are discussed at three levels. The principles of quasi-equivalence and the related geodesic dome formation are ®rst discussed conceptually and the geometric basis for their construction from two-dimensional assembly units is reviewed. The consequences for such an assembly when three-dimensional protein subunits are the associating components are then discussed with the coordinates of cowpea chlorotic mottle virus (CCMV) used to generate hypothetical structures in approximate agreement with the conceptual models presented in the ®rst section. Biophysical, molecular genetic, and atomic structural data for CCMV are then reviewed, related to each other, and incorporated into an assembly model for CCMV that is discussed with respect to the modular, chemical nature of the viral subunit structure. The concepts of quasi-equivalence are then examined in some larger virus structures containing multiple subunit types and auxiliary proteins and the need for additional control points in their assembly are considered. The conclusion suggests that some viral assembly principles are limited paradigms for protein associations occurring in the broader range of cell biology including signal transduction, interaction of transcription factors and protein traf®cking. # 1997 Academic Press Limited

*Corresponding author

Keywords: virus assembly; quasi-equivalence; structural polymorphism; cowpea chlorotic mottle virus; protein-protein interactions

Introduction The association of proteins controls many of the processes that are fundamental to biology. Various signal transduction pathways are dependent upon ligand-induced homo- and heteromeric protein multimerizations that result in a speci®c biological event. Such assemblies are characterized by ®nely tuned interaction energetics that respond properly to factors that determine protein association, dissociation and transmission of the signal to the next carrier in the pathway. The qualitative pathways of many of these processes have been mapped out, but only a few have been analyzed in atomic detail because of the dif®culty in isolating the components of such a system and studying them in vitro. Abbreviations used: CCMV, cowpea chlorotic mottle virus; BMV, bromegrass mosaic virus; ALMV, alfalfa mosaic virus; HK97, Hong Kong 97 bacteriophage; Nov, Nudaurelia capensis omega virus; ssRNA, singlestranded RNA; EM, electron microscopy. 0022±2836/97/250665±11 $25.00/0/mb971068

Icosahedral viral capsids are specialized protein assemblies that can, in many cases, be studied at atomic resolution by crystallography (e.g. see Speir et al., 1995). A number of viral capsids can be reversibly assembled in vitro from subunits puri®ed from disassembled virus particles (e.g. see Bancroft, 1970) or protein expressed in Escherichia coli or other systems, allowing the biophysical analysis of capsid assembly from wild-type or mutant subunits (e.g. see Fox et al., 1994, 1996; Zhao et al., 1995) under physiological and non-physiological conditions. Infectious clones are available for many simple plant and animal viruses and these allow the determination of a biological phenotype for those assembly mutants that retain infectivity (e.g. see Fox et al., 1997). Viral capsids composed of multiple copies of a single gene product and arranged with icosahedral symmetry provide a genetically economical way to generate an envelope that can package and transport the viral genome between cells of a susceptible host and between hosts. A common variety of viral particles, appropriate for investi# 1997 Academic Press Limited

666 gating the subtlety of protein-protein interactions, are capsids with quasi-equivalent symmetry. These particles display icosahedral symmetry, but are formed by more than 60 subunits, the largest number of proteins that can be assembled with each in an identical environment (e.g. see Johnson & Fisher, 1994). To assemble quasi-equivalent particles, the individual subunits and the subunit associations (morphological units) must display conformational polymorphism that depends on some type of molecular switching mechanism, a property shared with proteins involved in signal transduction (e.g. see Fanning & Anderson, 1996; Harrison, 1996). The geometric principles for quasi-equivalent assembly were developed by Caspar & Klug (1962). Since then a number of quasi-equivalent virus structures have been solved at moderate to high resolution by X-ray crystallography (reviewed by Johnson & Fisher, 1994) and, on the basis of these studies, an inventory of molecular switches has been identi®ed (Johnson, 1996). The switches invariably involve a segment (10 to 30 residues) of the subunit polypeptide that displays icosahedral order at some interfaces and is disordered at similar, but not symmetrically equivalent (quasi-equivalent), interfaces. Switches are normally multicomponent and include elements sensitive to the assembly environment. These may be divalent metal ions, hydrogen ions and single or doublestranded RNA in addition to the speci®c segments of the capsid polypeptide. In a sensitive and coordinated fashion these constituents alter quaternary structure interactions and result in closed protein shells formed of multiples of 60 copies of a single gene product that are capable of packaging and transporting the viral genome. Here, we will ®rst discuss the formal geometric requirements for the assembly of quasi-equivalent capsids identi®ed by Caspar and Klug more than 30 years ago, then graphically illustrate such an assembly process with subunit models derived from the X-ray structure of cowpea chlorotic mottle virus (CCMV; Speir et al., 1995). We then focus on the CCMV assembly system as it is understood from the structure, molecular genetics and biophysical studies. We will conclude by looking at quasi-equivalence in more complex viruses where auxiliary proteins, multiple subunit types, and preassembled modular intermediates are required to facilitate assembly. The Geometry of Quasi-equivalence The conceptual basis for viral quasi-equivalence is the interchangeable formation of hexamers and pentamers by the same protein molecule. A closed, quasi-equivalent, virus shell is geometrically equivalent to the geodesic domes of Buckminster Fuller (Fuller, 1975). Both are derived from a sheet of hexamers (Figure 1(a)) in which pentamers are inserted in place of certain hexamers according to selection rules described by a number T. With the

Quasi-equivalent Viruses: A Review

assignment of an origin, each hexamer in the sheet can be indexed according to the number of steps taken along each of the two axes (h and k) related by a 60 rotation. An equilateral triangle can be uniquely identi®ed for each hexamer indexed. A set of 20 such triangles can be arranged as shown in Figure 1(b) and these can be folded to generate a particle with exact icosahedral symmetry and the quasi-symmetry that corresponds to the indices of the hexamers chosen to de®ne the particular equilateral triangle. The hexagonal sheet is used as a formalism for constructing quasi-equivalent icosahedral lattices but is not strictly theoretical. The electron micrograph at the top right of Figure 1(a) shows that puri®ed CCMV coat protein, under speci®c conditions, can form two-dimensional crystalline arrays that are clearly composed of closepacked hexamers (lower image) observed after ®ltering and an increase in resolution (Adolph & Butler, 1974). Each triangular face of the icosahedron contains 3T subunits where T ˆ h2 ‡ hk ‡ k2 and h and k correspond to the indices for the hexamer chosen to de®ne the equilateral triangle. Figure 1(c) shows examples of particles with different T numbers and illustrates the increased volume for packaging the viral genome as the T number increases and the same size subunit is assumed. Each particle contains 20 triangular faces, 12 vertices, and 30 edges corresponding to the 3-fold, 5-fold and 2-fold rotational symmetry axes of the icosahedron. The size of the equilateral triangle is proportional to the T number and, assuming that the particles are formed of the same subunit type (gene product), each particle will contain 60T subunits (i.e. T ˆ the number of subunits in an icosahedral asymmetric unit). These will be arranged as 12 pentamers and 10(T ÿ 1) hexamers with their relative positions determined by the construction process described. The biological signi®cance of such a construction is evident when the close similarity in contacts required for the formation of the hexamer and pentamer are considered. Conceptually, there is only a subtle change in adjoining surfaces when a subunit of the hexamer is removed and the two subunits that were adjacent to it are brought into contact. Likewise, the contacts between hexameric and pentameric, and hexameric and hexameric morphological units, are also only subtly different because pentamers are pentavalent (surrounded by ®ve neighboring morphological units) and hexamers are hexavalent (surrounded by six neighboring morphological units). The importance of these close similarities and the switching required to accommodate the structural polymorphism is clear when we consider the assembly of a virus particle from asymmetric protein subunits that have thickness as well as the two-dimensional character of the hexagonal net. Figure 2 illustrates the conceptual basis for the formation of a ``biologically derived'' T ˆ 3 geodesic dome employing the Ca backbone of the CCMV subunit as the assembling module. The

Quasi-equivalent Viruses: A Review

667

Figure 1. Geometric principles for generating icosahedral quasi-equivalent surface lattices. These constructions show the relation between icosahedral symmetry axes and quasi-equivalent symmetry axes. The latter are symmetry elements that hold only in a local environment. (a) Hexamers are initially considered planar (an array of hexamers forms a ¯at sheet as shown) and pentamers are considered convex, introducing curvature in the sheet of hexamers when they are inserted. The closed icosahedral shell, composed of hexamers and pentamers, is generated by inserting 12 pentamers at appropriate positions in the hexamer net. To construct a model of a particular quasi-equivalent lattice, one face of an icosahedron is generated in the hexagonal net. The origin is replaced with a pentamer and the (h,k) hexamer is replaced by a pentamer. The third replaced hexamer is identi®ed by 3-fold symmetry (i.e. complete the equilateral triangle of the face). The icosahedral face for a T ˆ 3 surface lattice is de®ned by the triangle with bold lines (h ˆ 1, k ˆ 1). Two possible T ˆ 7 lattice choices are also marked with thin and dotted lines (h ˆ 2, k ˆ 1 or h ˆ 1, k ˆ 2, these being mirror images of each another), and require knowledge of the arrangement of hexamers and pentamers and the enantiomorph of the lattice for a complete lattice de®nition. Hexagonal sheets of CCMV coat protein have been viewed experimentally (top right, magni®cation 150,000  ; bottom right, top image with noise ®ltering and magni®cation 590,000  ; Adolph & Butler, 1974). For the purpose of these constructions, it is convenient to choose the icosahedral asymmetric unit as one-third of an icosahedral face de®ned by the triangle connecting a 3-fold axis to two adjacent 5-fold axes (other asymmetric units can be de®ned). (b) Seven hexamer units (bold outline in (a)) de®ned by the T ˆ 3 lattice choice are shown and the T ˆ 3 icosahedral face de®ned in (a) has been shaded. The icosahedral asymmetric unit is one-third of this face and it contains three quasi-equivalent units (two units from the hexamer coincident with the 3-fold axis and one unit from the pentamer). A three-dimensional model of the lattice can be generated by arranging 20 identical faces of the icosahedron as shown, and folded into a quasi-equivalent icosahedron. (c) Cardboard models of several icosahedral quasi-equivalent surface lattices constructed using the method described above. The procedure for generating quasi-equivalent models described here does not exactly correspond to that the described by Caspar & Klug (1962); however, the ®nal models are identical with those described in their paper.

hexameric sheet of subunits illustrated in parts (b) and (c) of the Figure is modeled on the basis of aberrant, but relevant, assembly products observed in vitro (Figure 1(a)) and was constructed by associating hexamers observed in the intact virus particle. All of the other molecular displays come directly from the near-atomic resolution structure of CCMV (Speir et al., 1995). All the subunits in hexamer sheets are in identical chemical environments; however, in an icosahedral arrangement, the three red and three green subunits can be related only by local 6-fold symmetry restricted to the boundaries of a single hexagonal polygon. True 6-fold symmetry (Figure 2(a) through (c)) is reduced to 3-fold symmetry (i.e. that of the white triangle, Figure 2(d) through (f)) with the insertion of pentamers and

relates the three pentamers to one another (blue subunits below the corners of the white triangle), the three hexamers to one another and, within the hexamers, the three red or the three green subunits to each other. Two levels of molecular switching are evident in the diagram. First, the subunits are capable of forming both hexamers and pentamers. Second, the pentamers are inserted among the hexamers in the appropriate locations. In the case of CCMV, ssRNA binding, protonation of carboxyl groups, and metal ion chelation across angular pentamer-hexamer and hexamer-hexamer interfaces (Figure 2(i)) stabilizes curvature in the absence of convex pentamers used to construct Figure 1. In the next section we discuss the in vitro studies of CCMV that provide a chemical basis for the assembly.

668

Quasi-equivalent Viruses: A Review

Figure 2. Molecular graphics construction of a T ˆ 3 quasi-equivalent icosahedron. (a) Hexagonal sheet overlaid with the triangular coordinates (white) for a theoretical T ˆ 3 quasi-equivalent icosahedron (h ˆ 1, k ˆ 1, see Figure 1(a) and (b)). The sheet has true 6-fold rotational symmetry about axes passing through the hexamer centers that are normal to the sheet. (b) Copies of the hexamer coordinates from the CCMV X-ray structure can be positioned in the sheet by simple translations. (c) A sideview of the modeled sheet demonstrates its planarity. (d) Hexamers at the corners of the white (h ˆ 1, k ˆ 1) triangle become pentamers. The planar sheet (yellow model) takes on curvature to maintain contacts between the polygons (green model). (e) The magnitude of the pentamer-induced curvature is displayed in the side-view of the partial polyhedron. (f) Coordinates of the CCMV X-ray structure ®t this construction without any manipulation. (g) A completed T ˆ 3 icosahedral model. The 12 pentamers generate curvature that closes the structure. This cage (a truncated icosahedron) accurately describes the geometric morphology of CCMV (h) which is composed of modular, planar, pentamers (12) and hexamers (20). Angular pentamer-hexamer and hexamer-hexamer interfaces (i) stabilize curvature in the absence of convex pentamers used to construct Figure 1.

The Chemical Basis for Quasiequivalence: Structural Polymorphism in CCMV The CCMV subunit provides the basis for tertiary and quaternary structure polymorphism. Figure 3(a) is a ribbon drawing of the subunit and the regions highlighted indicate portions that control quaternary structure. CCMV subunits are composed of 190 residues folded in a canonical viral b-sandwich (Rossman & Johnson, 1989). Residues 27 to 42 are ordered when the subunits form hexamers (b-hexamer, Figure 3(b)) and they are not visible in the electron density (the residues lack icosahedral symmetry for this portion of the polypeptide) when the subunits form pentamers, indicating a role for these residues in modulating the stability of these two multimers. The b-hexamer is a striking example of quasi-symmetry where the residues of the six separate polypeptide strands obey nearly perfect local 6-fold symmetry coincident with an icosahedral 3-fold symmetry axis. Stability is derived from numerous interactions within and around the b-hexamer structure, such as main-chain hydrogen bonds and hydrophobic side-chain interactions. Residues 180 to 190 near the C terminus of the subunit interact with neighboring, 2-fold related, subunits to form the soluble

and assembling dimer unit. These residues must permit ¯exibility to allow for the different dimer interactions in the quasi-equivalent particle and they play a crucial role in permitting the proper assembly to be achieved. Following coat protein dimerization, the b-hexamer structure is likely to be the ®rst product of virion assembly (see Figure 5). The regions highlighted in the body of the protein contain residues sensitive to pH, ionic strength and the presence of metal ions or ssRNA. These regions determine the pathway of assembly and, indirectly, determine the formation of hexamers and pentamers as the subunits associate. The consequences of these functional regions on assembly are shown in Figure 4, which was adapted from a review article by Bancroft (1970). Although complex, this assembly phase diagram clearly illustrates the balance of interactions required to drive proper particle formation. The native CCMV virion (Figures 2(h) and 4) has divalent metal ions associated with it and has a velocity sedimentation coef®cient of 85 S. Metal ions can be removed at pH 5.0 with EDTA, generating particles that are virtually indistinguishable from native particles. If, however, metal ions are removed at pH 7 or, if metal-free particles at pH 5 are transferred to a buffer at pH 7, the swollen form of the capsid results (Figure 4). Bancroft

Quasi-equivalent Viruses: A Review

Figure 3. (a) Modular character of the CCMV coat protein. The view is approximately tangential to the particle curvature. The interior and exterior of the virus particle are labeled along with the functions of the highlighted regions. (b) The CCMV b-hexamer. The view is identical with (a). The labels B2, C and B5 denote the relative placement of the subunits in the complete virus shell and are used in Figure 5.

(1970) correctly interpreted this as the result of electrostatic repulsion of charged carboxyl clusters (depicted in the diagram) that exist in the absence of divalent metal ions at pH 7.0 but are protonated at pH 5.0. The clustering of these residues results in anomalously high pK values for these carboxylate groups similar to what has been observed in tobacco mosaic virus (Namba et al., 1989). These residues are highlighted in the b-sandwich in Figure 3(a). The swollen form of the virus is susceptible to nuclease digestion and can be disassembled into protein and nucleic acid by raising

669 the ionic strength to 1 M NaCl. It is clearly stabilized primarily by protein-nucleic interactions and residues responsible for this stability are identi®ed in Figure 3(a). When these ionic interactions have been neutralized by salt, the subunits exist as dimers in solution. Reassembly into infectious particles is straightforward and highly ef®cient by mixing protein and nucleic acid, lowering the ionic strength to 0.02 M, maintaining the pH at 7.0 and adding divalent metal ions. If the reassembly is performed at pH 5.5 with or without metal ions, a slightly slower sedimenting particle results. The protein alone will reassemble into a variety of polymorphic forms that are highly sensitive to ionic strength and pH. All of the forms at pH 7 and moderate ionic strength are tubes, sheets and multishelled particles. At pH 5.0 and moderate ionic strength, protein shells with T ˆ 3 quasi-symmetry form. At lower pH and ionic strength, laminar structures form. All of these assembles are explainable in terms of the interactions observed in the native particles and suggest the likely assembly pathway of the virus. Figure 5 shows a structure-based pathway of assembly for the virions that is consistent with the in vitro studies described. Soluble, non-assembled, subunits exist as dimers and have structural characteristics that identify them with the dimers formed in an assembled CCMV particle (Zhao et al., 1995). The dimers are ¯exible about their connection and they can exist in either a ¯at or bent con®guration as shown at the top of Figure 5. At neutral pH and low ionic strength in the absence of RNA and metal ions, the subunits form ¯at dimers. They initially assemble as a hexamer of dimers (Figure 5(b)) because of the added stability of the b-hexamer (Figure 3(b)). Subunits in hexamers have twice the number of contacts as in pentamers (second only to subunit dimer contacts) due to the formation of the b-hexamer. Only dimers can add to the 12mer (two 12mers cannot combine) and without inducement of curvature the assembly will grow outward from this nucleation oliogomer as hexamer sheets (Figure 1(a)) and all hexamer tubes having large radii of curvature (Figure 4). bHexamer formation will localize six N-terminal basic regions to aid in neutralizing the acidity of the ssRNA and condensing the genome. The interface between polygons A and C becomes one of six possible ssRNA and calcium-binding sites created by 12mer formation. If assembly occurs at pH 5.0, or at neutral pH in the presence of divalent metal ions and RNA, curvature is induced and it becomes energetically favorable to add pentamers to minimize the total surface area of the closed particle. Pentamers in solution are improbable due to the required position of the N terminus to create dimer contacts (Speir et al., 1995) and several assembly studies have shown in the absence of nucleic acid and divalent metal ions, that there is a tendency to form hexamer-rich structures with little or no curvature. It is likely that pentamers form only during the assembly process or as needed to

670

Quasi-equivalent Viruses: A Review

Figure 4. An assembly ``phase diagram'' for CCMV derived from in vitro assembly studies (adapted from Bancroft, 1970).

maintain inter-subunit contacts where curvature has occurred. Proper quasi-equivalent assembly occurs when the relative stabilities of hexamers and pentamers are in the proper ratio. In isolation, hexamers are intrinsically more stable than pentamers because of the added interactions of the b-hexamer, but it is unlikely that either pentamers or hexamers exist in isolation during CCMV assembly once the initial nucleation has begun. Instead, particle formation is always occurring at the growing edge of a sheet and here the curvature discussed above will allow more area of surface to interact when pentamers are inserted into the proper locations. The total buried surface area for 180 CCMV subunits modeled as a hexamer sheet (Figure 2(b)) is Ê 2, while the buried surface area for the 632  103 A Ê 2, properly assembled T ˆ 3 particle is 717  103 A indicating that the latter is likely to have the greater stability and accounts for the formation of pentamers in particle assembly. The importance of the relative stabilities of hexagonal and pentagonal morphological units during assembly was recently illustrated with experimental data for brome mosaic virus (BMV), a bromovirus with signi®cant sequence identity with CCMV (Flasinski et al., 1997). A mutant was discovered with greatly decreased virulence. A single residue on the bF-bG loop was identi®ed as the mutation (Figure 3(a)). This residue contributes signi®cantly to the formation of hexamers and pentamers in CCMV and, by analogy, the replacement appeared likely to increase the stability of the hex-

amer in BMV, thereby altering the relative stabilities of hexamers and pentamers, and interfering with proper assembly. Analysis of the small yield of particles produced by this mutant con®rmed that they were less stable than wild-type particles. The hypothesis of an assembly defect was supported when a number of second sight mutations were identi®ed that restored nearly wild-type virulence. All of these changes were located in the CCMV model and appeared to compensate for the original mutation by destabilizing the hexamer, thus restoring the proper ratio of hexamer-pentamer stability. Some plant viruses have evolved to over-stabilize hexamers relative to pentamers resulting in the formation of bacilliform particles. Alfalfa mosaic virus (ALMV) is an example where hexamers are the dominant morphological form and strong protein-RNA interactions allow the size of the RNA molecule to determine the particle size (Mellema & van den Berg, 1974). Hexamers are organized on a cylindrical lattice and only at the two ends are pentamer interactions required to close the tube. Although ALMV capsid protein preferentially packages its own genomic components, in the absence of ALMV RNA, the subunits will indiscriminately package RNA or DNA of virtually any size, in vitro, and form a bacilliform particle of a proportional length (Hull, 1970). A similar variation on the theme of quasi-equivalence may occur in retroviruses where protein-RNA complexes of polymorphic forms have been assembled in vitro (Campbell & Vogt, 1995) and polymorphism may

Quasi-equivalent Viruses: A Review

671

Figure 5. Proposed assembly pathway for the T ˆ 3 CCMV particle. (a) CCMV subunits exist as dimers in solution (Zhao et al., 1995). Here a single subunit is drawn as a foursided polygon shaped like an arrow. Two of these are shown related by a 2-fold symmetry axis normal to the page and passing through the point where the corners of these polygons are touching. (b) Particle assembly begins with the formation of a hexamer composed of six dimers (12mer) either in vivo or in vitro under the conditions shown. The 12mers may exist in equilibrium with dimers (heavily favoring dimers). Four of the polygons (two dimers: C and C2, A and B5; subscripts refer to the x-fold rotational symmetry operation used to generate the subunit from the original A-B-C polygons of the icosahedral asymmetric unit) have been colored and labeled as to their ®nal con®guration in (e) to aid in their identi®cation through the remaining steps. The 12mer structure is planar and reminiscent of the hexagonal sheets shown in Figures 1(a) and 2(a) through (c). The ``¯at'' dimer contacts are shown diagrammatically above the 12mer and their views have speci®c orientations (refer to (e)) shown by the symmetry symbols above them (white, quasi-icosahedral symmetry; yellow, icosahedral symmetry; oval, 2-fold; triangle, 3-fold; pentagon, 5-fold). (c) Icosahedral assembly occurs rapidly in the presence of the ssRNA genome and divalent metal cations. The CCMV particle morphology is accurately displayed in (e) as a truncated icosahedron (12 planar pentamers and 20 planar hexamers) and curvature results from nearly equal angular subunit interactions within the icosahedral asymmetric unit at the interfaces between planar pentamer and hexamer units (Figure 2(i)). These interfaces can bind both calcium (brown circles) and ssRNA (shown above (c), not shown below, since it is located directly beneath the calcium sites), and each additional dimer (shown in gray) creates two new binding sites. (d) Curvature induced by ligand binding drives the formation of pentamers. The morphology of the T ˆ 3 icosahedral shell appears to be designed around ssRNA and calcium ion binding, which outlines the pentamer and hexamer boundaries. (e) Rapid addition of 84 dimers to the 12mer completes the particle and creates 180 potential ssRNA and calcium binding sites. CCMV virions are sensitive to pH and ionic strength in the surrounding environment. Using the conditions shown they can be rapidly disassembled in vitro back to dimers.

be an essential part of the particle maturation process.

Limits of Quasi-equivalent Shell Formation In theory, quasi-equivalent shells of any size can be formed. In practice, the largest shells formed of a single subunit type are T ˆ 7, although shells with much larger T numbers are found if the capsid contains more than one subunit type and/or if the capsid is assembled from preassembled modular units. Capsids with large T numbers appear to require one or more auxiliary proteins to control proper assembly and may utilize different subunit types to form hexamers and pentamers. These observations suggest a limit to the size of a shell that can be reliably assembled by controlling only stability ratios of hexamers and pentamers formed by a single subunit type. The auxiliary proteins have different functions in different viruses and may be involved at various

stages of assembly. Double-stranded DNA phage such as P22, f29, or the very large T4 phage, utilize a scaffold protein that guides the association of subunits into a closed shell and is released and recycled during or after proper assembly (reviewed by Casjens & Hendrix, 1988). In contrast members of the herpes and adeno virus families have minor proteins that remain associated with the capsids and they may direct the relative positioning of hexamers and pentamers during assembly or stabilize their structures post assembly (e.g. see Stewart et al., 1993). In some cases the ``auxiliary'' protein may be a domain encoded in viral subunits that form large T numbers. This domain functions as a more elaborate version of the small polypeptide molecular switches that are encoded within the single subunits of the viruses with smaller T numbers. Figure 6 illustrates virus capsids with T numbers between 3 and 25, that have been determined with cryo-electron microscopy and image reconstruction. These structures show that the surface lattices (the disposition of hexamers and pentamers depicted in the models in Figure 1) predicted by

672

Quasi-equivalent Viruses: A Review

Figure 6. Cryo-electron micrograph reconstructions of viruses with T values between 3 and 25. In each case the equivalent of one face of an icosahedron is outlined (the vertices of the triangle are at 5-fold symmetry axes and an icosahedral 3-fold symmetry axis is at the center of the triangle, icosahedral 2-fold symmetry axes lie midway along the lines connecting 5-fold axes). The outlined faces for each virus can be compared with the distribution of hexamers and pentamers on faces of the quasi-equivalent models in Figure 1(c). Examples of both enantiomorphs of the T ˆ 7 lattice are shown (the hexavalent lattice sites within the outlined icosahedral faces are mirror images) as diagrammed Ê to the T ˆ 25 adenovirus in Figure 1(a). The viruses range in size from the T ˆ 3 CCMV with a diameter of 285 A Ê (not including the spikes). All the particles are shown with their exterior surfaces except with a diameter of 1000 A Ê where the T ˆ 13l symmetry is most apparent. rotavirus, which has been radially truncated to a diameter of 670 A Nov is the only T ˆ 4 virus structure determined at near atomic resolution (Johnson et al., 1994; Munshi et al., 1996).

quasi-equivalence are followed with high ®delity; however, the detailed oliogomeric structures occupying the hexamer positions may be true hexamers, distorted hexamers, pseudo-hexamers or, in one case, pentamers. Recognizing that the hexamer symmetry depicted by the models is an idealization that is not frequently observed in virus structures, it may be better to refer to the ``hexamer'' positions as hexavalent positions indicating that an oligomer occupying one of these sites will always have interactions with six neighboring oligomers. This is generally correct and describes the functionally important feature of oligomers occupying the site; they must be able to interact with six neighboring oligomeric units in such a way that the surface lattice is preserved. In the paragraphs below, a number of virus structures are described with T numbers greater than 3 and the nature of the polymorphism or multiple subunit interactions required to propagate the particular lattice are discussed. Assembly of the T ˆ 7, temperate, l-like phage HK97 is probably an example of an elaborate molecular switch being encoded as part of a single subunit. Unlike all other dsDNA phage investigated, HK97 assembles without the bene®t of separate scaffolding protein. The capsid subunit is a 41 kDa gene product that assembles into a T ˆ 7l capsid (Figure 6). There are six copies of the subunit occupying the hexavalent lattice points, but they do not have 6-fold symmetry. The ``hexamers'' appear as half hexagons that have been translated relative to each other, generating a dimer of apparent trimers (Figure 6). A virally encoded

protease is packaged in the particle and it digests 102 amino acid residues from the N terminus following assembly. The protease also digests itself and the digestion products diffuse out of the capsid. The association of full-length subunits made in a bacterial expression system can be studied in vitro and they will form an equilibrium mixture of hexamers and pentamers. The position of this equilibrium can be arti®cially shifted by altering the solution conditions (Xie & Hendrix, 1995). Following assembly, cleavage of the subunits and initial packaging of DNA the particle undergoes a dramatic maturation that includes an expansion of the Ê to 570 A Ê shell from an average diameter of 450 A and a covalent crosslinking of the subunits (Conway et al., 1995; Duda et al., 1995a,b). These observations suggest that the ®rst 102 residues of the capsid protein are a covalently associated switch or scaffold that is used and then discarded through digestion rather than functioning catalytically and reusably as in P22 (King & Casjens, 1974). Indeed the 102 residue
-region, as it is called, has a sequence of heptad repeats suggestive of an a-helical structure designed for self-association and this may explain the initial assembly of hexamer and pentamer units described above (Conway et al., 1995). The organization of these morphological units in the procapsid may be driven by their relative stabilities as described for the CCMV particles. Rhesus rotavirus is an example of a complex capsid made of multiple gene products and multiple layers that conform to T ˆ 13l quasi-symmetry (Prasad et al., 1988, 1990; Yeager et al., 1990, 1994). Quasi-symmetry breaks down at the outer

673

Quasi-equivalent Viruses: A Review

most part of the particle where 60 dimeric copies Ê above the of protein VP4 (88 kDa) extend 100 A outer quasi-equivalent shell. The outer T ˆ 13l shell has inner and outer radial dimensions of Ê and 392 A Ê , respectively, and is formed by 365 A 780 copies of protein VP7 (34 kDa). The inner T ˆ 13l shell, which is in register with the outer shell, has inner and outer radial limits of approxiÊ and 365 A Ê , respectively, and is commately 300 A posed of 780 copies of protein VP6 organized as 260 trimeric columns. Throughout the shells there are open channels that provide access from inside the shell to the external surface. Figure 6 shows the obvious T ˆ 13l lattice formed by subunits VP6 when the density is displayed as a radial surface at Ê . The complexity of this virus and the lack of 335 A experimental assembly studies precludes a proposal for the mechanism of assembly. It is clear, however, that the energetics of quasi-equivalence govern the formation of the majority of the protein shell. Ê diameter without Adenovirus, a large (1000 A the ®bers), non-enveloped DNA virus, demonstrates another novel strategy for achieving large T numbers, actually pseudo T numbers, since a T number formally applies only to capsids made of a single subunit type. The adenovirus capsid is composed of a variety of proteins that have been mapped out in position, stoichiometry and, for one of the proteins, atomic resolution detail by Burnett and his collaborators (Athappilly et al., 1994; Burnett, 1985; Burnett et al., 1985; Stewart et al., 1991, 1993). The hexon subunits (subunits that form pseudo-hexamers) and penton base subunits (subunits forming true pentamers) create the adenovirus surface with 12 pentamers on the vertices of the icosahedron and 240 hexons (pseudo hexamers) covering the faces. Without further analysis this apparently adds to 1500 subunits and corresponds formally to a T ˆ 25 surface lattice if all the subunits were identical (Figure 6). The polymorphism required to generate such a pseudo surface lattice is novel and utilizes less than the 25 different environments suggested by the pseudo T number. First, the ``hexamer'' is actually a trimer! Three identical subunits associate to form an oligomer with pseudo 6-fold symmetry at its base and trimeric symmetry at its outer surface Organization of the trimer suggests that it forms as a preassembly unit, since there is extensive interchain association of the subunits dependent on oligomerization. Burnett points out that considering the hexon-trimer as the basic unit of the shell results in the hexon-trimer forming only four unique contacts with neighbors rather than the 25 separate interactions suggested by the pseudo T number. Penton base subunits occupy another environment, but this is effected through the use of a different polypeptide chain. It is clear, however, that the original premise of quasi-equivalence survives in part. The pseudo 6-fold symmetry of the trimer, generated by 3-fold related sets of two, pseudoequivalent, b-sandwich structures within each sub-

unit polypeptide chain, allows the hexon to satisfy the hexavalent lattice sites that it occupies on the classical T ˆ 25 quasi-equivalent lattice. Structural and biochemical studies have shown that a speci®c ``group of nine'' of hexon-trimers are a stable unit that is ``fastened'' together with an auxiliary protein (polypeptide IX) probably after assembly. Other auxiliary proteins differentially stabilize contacts between hexons, penton bases and the capsid with the nucleoprotein core. Hence, adenovirus achieves a pseudo, quasi-equivalent T ˆ 25 shell by making a variety of approximations to the original concepts proposed, yet the result is a lattice that places pentamers and pseudo hexamers on the rotation and quasi-rotational axes predicted by Caspar & Klug (1962). These examples illustrate that there is an intrinsic minimal energy associated with the quasiequivalent surface lattice and that the lattice is utilized even when the original hypothesis for its existence, the presence of a single type of subunit in the capsid that is capable of forming hexamers and pentamers, is violated. The most novel example of pseudo quasi-equivalence is the case of papovavirus capsids in which the subunits can associate only as pentameric morphological units (polyoma virus, Figure 6; Baker et al., 1983; Rayment et al., 1982). In spite of this restriction, which suggests that these subunits can form only a T ˆ 1 shell, the all pentamer T ˆ 7 surface lattice is observed (Liddington et al., 1991; Stehle et al., 1994). This entirely violates the original basis for quasi-equivalence, that hexamers occupy hexavalent surface lattice sites and pentamers occupy pentavalent lattice sites. Instead, pentamers occupy both hexameric and pentameric lattice sites, creating extraordinary mismatches in subunit associations that are accommodated by a switching mechanism controlled by the C-terminal portion of the subunit. Rather than switching between hexamers and pentamers with the relative positioning of these units determined by their relative stabilities, there are six different switch states found in the high-resolution structure of the capsid. In spite of extraordinary approximations, papovaviruses maintain the surface lattice predicted by quasiequivalent theory.

Conclusions The discussion above demonstrates that there are many ways to generate the variable interactions between subunits required for the formation of quasi-equivalent virus shells. As each high-resolution structure of a virus with a quasiequivalent or pseudo quasi-equivalent shell has been determined, a variation of molecular switching mechanisms or other strategies for achieving polymorphism have been identi®ed. Quasi-equivalence is a remarkable example of the productive interplay between theory and experiment. Caspar & Klug (1962) provided a reliable basis for under-

674 standing minimal free-energy structures of viruses, but only empirical analysis has demonstrated the versatility of the nucleoprotein complexes required to achieve the chemical interactions to produce such structures. It is reasonable to anticipate that a variety of mechanisms will be found as the study of oligomerization domains involved in other homo- and heteromeric subunit associations advances. Recently, Harrison (1996) has examined the type of oligomerization that occurs in proteins that associate in a signal transduction pathway. In a number of cases, protein associations occur through a mechanism similar to that described for CCMV. In the environment conducive for association of two proteins, a polypeptide portion of one protein will have a well-de®ned structure and interact with the surface of the correlate protein. Other environments lead to multiple conformations of the interacting polypeptide and no association. As Harrison (1996) points out, similar balanced associations are likely to occur in transcriptional initiation and traf®cking of proteins. Thus, as was the case in the development of other areas of biophysics, virus structure provides a microcosm of biology that can be readily investigated and related to other areas of biology.

Acknowledgements We thank Professor Mark Young, Dr James Fox, Dr Sanjeev Munshi and Ms Xiaoxiao Zhao for many helpful discussions on the assembly of CCMV; Professor Andrew Fisher for his help in preparing parts of Figure 1; and Ms Sangita Sinha for constructing the models in Figure 1(c). The virus structures in Figure 6 were kindly provided by Professor Roger Burnett and Dr John Rux (adenovirus), Professor Mark Yeager and Mr Brian Sheehan (rotavirus), Dr James Conway (HK97), and Professor Tim Baker and Mr Norm Olson (CCMV, Nov, polyoma). This work was supported by National Institutes of Health grant GM54076 to J.E.J. J.A.S. was supported in part by National Institutes of Health Neurosciences training grant MH19185

References Adolph, K. W. & Butler, P. J. G. (1974). Studies on the assembly of a spherical plant virus. I. States of aggregation of the isolated protein. J. Mol. Biol. 88, 327 ± 341. Athappilly, F. K., Murali, R., Rux, J. J., Cai, Z. & Burnett, R. M. (1994). The re®ned crystal structure of hexon, the major coat protein of adenovirus type Ê resolution. J. Mol. Biol. 242, 430 ± 455. 2, at 2.9 A Baker, T. S., Caspar, D. L. & Murakami, W. T. (1983). Polyoma virus `hexamer' tubes consist of paired pentamers. Nature, 303, 446 ± 448. Bancroft, J. B. (1970). The self-assembly of spherical plant viruses. Advan. Virus Res. 16, 99 ±134. Burnett, R. M. (1985). The structure of the adenovirus capsid. II. The packing symmetry of hexon and its

Quasi-equivalent Viruses: A Review implications for viral architecture. J. Mol. Biol. 185, 125± 143. Burnett, R. M., Grutter, M. G. & White, J. L. (1985). The structure of the adenovirus capsid. I. An envelope Ê resolution. J. Mol. Biol. 185, model of hexon at 6 A 105± 123. Campbell, S. & Vogt, V. M. (1995). Self-assembly in vitro of puri®ed CA-NC proteins from Rous sarcoma virus and human immunode®ciency virus type 1. J. Virol. 69, 6487± 6497. Casjens, S. & Hendrix, R. (1988). Control mechanisms in dsDNA bacteriophage assembly. In The Bacteriophages (Calendar, R., ed.), vol. 1, pp. 15± 91, Plenum Publishing Corp. Caspar, D. L. D. & Klug, A. (1962). Physical principles in the construction of regular viruses. Cold Spring Harbor Symp. Quant. Biol. 27, 1 ± 24. Conway, J. F., Duda, R. L., Cheng, N., Hendrix, R. W. & Steven, A. C. (1995). Proteolytic and conformational control of virus capsid maturation: the bacteriophage HK97 system. J. Mol. Biol. 253, 86±99. Duda, R. L., Hempel, J., Michel, H., Shabanowitz, J., Hunt, D. & Hendrix, R. W. (1995a). Structural transitions during bacteriophage HK97 head assembly. J. Mol. Biol. 247, 618± 635. Duda, R. L., Martincic, K. & Hendrix, R. W. (1995b). Genetic basis of bacteriophage HK97 prohead assembly. J. Mol. Biol. 247, 636± 647. Fanning, A. S. & Anderson, J. M. (1996). Protein-protein interactions: PDZ domain networks. Curr. Biol. 6, 1385± 1388. Flasinski, S., Dzianott, A., Speir, J. A., Johnson, J. E. & Bujarski, J. J. (1997). Structure-based rationale for the rescue of systemic movement of brome mosaic virus by spontaneous second-site mutations in the coat protein gene. J. Virol. 71, 2500 ±2504. Fox, J. M., Johnson, J. E. & Young, M. J. (1994). RNA/ protein interactions in icosahedral virus assembly. Sem. Virol. 5, 51± 60. Fox, J. M., Xiaoxia, Z., Speir, J. A. & Young, M. J. (1996). Analysis of a salt stable mutant of cowpea chlorotic mottle virus. Virology, 222, 115± 122. Fox, J. M., Albert, F. G., Speir, J. A. & Young, M. J. (1997). Characterization of a disassembly de®cient mutant of cowpea chlorotic mottle virus. Virology, 227, 229± 233. Fuller, R. B. (1975). Synergetics, pp. 314±429, Macmillan, New York. Harrison, S. C. (1996). Peptide-surface association: the case of PDZ and PTB domains. Cell, 86, 341± 343. Hull, R. (1970). Studies on alfalfa mosaic virus. III. Reversible dissociation and reconstitution studies. Virology, 40, 34± 47. Johnson, J. E. (1996). Functional implications of proteinprotein interactions in icosahedral viruses. Proc. Natl Acad. Sci. USA, 93, 27± 33. Johnson, J. E. & Fisher, A. J. (1994). Principles in virus structure. In Encyclopedia of Virology (Webster, R. G. & Granoff, A., eds), pp. 1573± 1586, Academic Press, London. Johnson, J. E., Munshi, S., Liljas, L., Agrawal, D., Olson, N. H., Reddy, V., Fisher, A., McKinney, B., Schmidt, T. & Baker, T. S. (1994). Comparative studies of T ˆ 3 and T ˆ 4 icosahedral RNA insect viruses. Arch. Virol. Suppl. 9, 497± 512. King, J. & Casjens, S. (1974). Catalytic head assembling protein in virus morphogenesis. Nature, 251, 112 ± 119.

Quasi-equivalent Viruses: A Review Liddington, R. C., Yan, Y., Moulai, J., Sahli, R., Benjamin, T. L. & Harrison, S. C. (1991). Structure Ê resolution. Nature, 354, of simian virus 40 at 3.8 A 278± 284. Mellema, J. E. & van den Berg, H. J. N. (1974). The quaternary structure of alfalfa mosaic virus. J. Supramol. Struct. 2, 17± 31. Munshi, S., Liljas, L., Caverelli, J., Bomu, W., McKinney, Ê B., Reddy, V. & Johnson, J. E. (1996). The 2.8 A structure of a T ˆ 4 animal virus and its implications for membrane translocation of RNA. J. Mol. Biol. 261, 1 ±10. Namba, K., Pattanayek, R. & Stubbs, G. (1989). Visualization of protein-nucleic acid interactions in a virus. Re®ned structure of intact tobacco mosaic virus at Ê resolution by X-ray ®ber diffraction. J. Mol. 2.9 A Biol. 208, 307 ± 325. Prasad, B. V. V., Wang, G. J., Clerx, J. P. M. & Chiu, W. (1988). Three-dimensional structure of rotavirus. J. Mol. Biol. 199, 269±275. Prasad, B. V. V., Burns, J. W., Marietta, E., Estes, M. K. & Chiu, W. (1990). Localization of VP4 neutralization sites in rotavirus by three dimensional cryo-electron microscopy. Nature, 343, 476± 479. Rayment, I., Baker, T. S., Caspar, D. L. & Murakami, W. T. (1982). Polyoma virus capsid structure at Ê resolution. Nature, 295, 110 ±115. 22.5 A Rossmann, M. G. & Johnson, J. E. (1989). Icosahedral RNA virus structure. Annu. Rev. Biochem. 58, 533 ± 573. Speir, J. A., Munshi, S., Wang, G., Baker, T. S. & Johnson, J. E. (1995). Structures of the native and swollen forms of cowpea chlorotic mottle virus

675 determined by X-ray crystallography and cryo-electron microscopy. Structure, 3, 63 ± 78. Stehle, T., Yan, Y., Benjamin, T. L. & Harrison, S. C. (1994). Structure of murine polyomavirus complexed with an oligosaccharide receptor fragment. Nature, 369, 160 ± 163. Stewart, P. L., Burnett, R. M., Cyrklaff, M. & Fuller, S. D. (1991). Image reconstruction reveals the complex molecular organization of adenovirus. Cell, 67, 145 ± 154. Stewart, P. L., Fuller, S. D. & Burnett, R. M. (1993). Difference imaging of adenovirus: bridging the resolution gap between X-ray crystallography and electron microscopy. EMBO J. 12, 2589± 2599. Xie, Z. & Hendrix, R. W. (1995). Assembly in vitro of bacteriophage HK97 proheads. J. Mol. Biol. 253, 74 ± 85. Yeager, M., Dryden, K. A., Olson, N. H., Greenberg, H. B. & Baker, T. S. (1990). Three-dimensional structure of rhesus rotavirus by cryoelectron microscopy and image reconstruction. J. Cell Biol. 110, 2133± 2144. Yeager, M., Berriman, J. A., Baker, T. S. & Bellamy, A. R. (1994). Three-dimensional structure of the rotavirus haemagglutinin VP4 by cryo- electron microscopy and difference map analysis. EMBO J. 13, 1011± 1018. Zhao, X., Fox, J. M., Olson, N. H., Baker, T. S. & Young, M. J. (1995). In vitro assembly of cowpea chlorotic mottle virus from coat protein expressed in Escherichia coli and in vitro-transcribed viral cDNA. Virology, 207, 486± 494.

Edited by T. Richmond (Received 8 January 1997; received in revised form 24 February 1997; accepted 26 February 1997)