Phi29 assembly intermediates reveal how scaffold interactions with capsid protein drive capsid construction and maturation

Abstract

The self-assembly of bacteriophage capsids from major capsid proteins (MCPs) and scaffolding proteins (SPs) and the subsequent expansion of these capsids are essential steps in bacteriophage life cycles. However, the mechanism by which assembly occurs remains poorly understood, and few intermediate states are available to illuminate the expansion of meta-stable procapsids into robust mature capsids. Here, we present the structure of a partially expanded phi29 procapsid that reveals distinct conformations of MCPs and allows visualization of SPs in multiple oligomeric states. These results suggest that formation of SP dimers, tetramers, and higher-order oligomers drives dissociation of SP from MCP to actuate capsid expansion. Hexons expand first, and we propose penton maturation is delayed by a symmetry match with SP oligomers. We further show that the prolate shape of phi29’s capsid is possible due to concave hexons in the equatorial region of the capsid that may alter interactions with SP and explain the observed dependence of the prolate shape on SP.

A bacteriophage phi29 capsid maturation intermediate shows how scaffold and capsid protein direct capsid assembly and maturation.

INTRODUCTION

Research into bacteriophages and their interactions with host cells has been a cornerstone of molecular biology, uncovering foundational principles of life at the molecular level (1–4). Tailed phage are ubiquitous and highly diverse, and thus, the shells that encapsidate and protect phage genomes, made from virally encoded major capsid proteins (MCPs), occur in a wide range of shapes and sizes. These MCPs represent a treasure trove of structural knowledge; an enormous sequence space has been explored over ~3.5 billion years of viral evolution among a 10³¹ population size. Further bacteriophage capsids have proven amenable to structural analysis by cryo–transmission electron microscopy (cryo-TEM), and the natural selective pressures that have guided their evolution have been highly studied (5). Hence, analysis of phage MCP structures can illuminate how protein structure and evolution relate to function. In particular, evolution selects MCPs based on their ability to perform at least three critical functions: (i) self-assembly—polymerizing from independent soluble subunits, possibly around a nucleating structure, into a specific closed geometry while avoiding a wide range of potential open structures; (ii) quasi-equivalence (QE)—the ability of subunits of the same sequence to structurally adapt to different geometric requirements present at different regions of the shell (6); and (iii) maturation—coordinated conformational transformation of all the monomers in a capsid into the highly stable mature form of the virus. Each of these three functions is facilitated by the temporary presence of “scaffolding proteins” (SPs), which have thus far been largely resistant to structural characterization. Hence, although SPs play a crucial role in capsid construction and maturation, the mechanisms by which they do so remain elusive.

Despite billions of years of evolution resulting in structural proteins with highly diverse sequences, tailed phage MCPs are remarkably structurally homologous. MCP adopts a fold named for HK97, the phage in which it was first described (7, 8). The HK97 fold consists of four canonical structural domains, the A- and P-domains, the E-loop, and the N-arm, each of which contributes differently to the MCP functions described earlier (Fig. 1). The wedge-shaped A-domain occupies the axial center of the capsomers. Its loops and helices adopt different conformations at different quasi-equivalent positions and before and after prohead maturation. The long plank-shaped P-domain flexes to provide a range of curvature adapted to local QE environments. The E-loop extends along the capsomer perimeter, in the opposite direction of the P-domain. Each E-loop binds to the P-tip of an adjacent capsomer to create a network of interpenetrating rings that encompasses the entire capsid (9). The N-arm is, as we shall show, the control center for maturation of the HK97 fold. It completely refolds during the course of prohead maturation, going from an α helix in the interior of the prohead to an extended state that forms an intermolecular β sheet with the E-loop of a neighboring monomer and reaches the exterior of the mature capsid. The prohead-state N-arm has been shown to be responsible for binding the SP to the MCP in several phage (10). Similarly, in other phages, the SP and MCP are initially expressed as a single contiguous polypeptide that is later cleaved at the beginning of the MCP N-arm. During capsid assembly, MCP must be prevented from switching to the stable, essentially irreversible mature state, while monomers may still need to rearrange into the correct QE ordering to yield a stable closed shell. SP is known to play a role in this delay of maturation. Its dissociation from the N-arm, either upon entry of DNA into the prohead or upon protease cleavage in cases where SP is initially expressed as a contiguous Δ-domain of the MCP, has been shown to be the trigger for prohead maturation for several species of phage (10, 11). Maturation proceeds as a rapid chain reaction (12); while some of the most stable intermediate states have been observed, more reactive states are too short-lived for structural analysis. Mutations or unusual conditions that stabilize these elusive intermediates could provide valuable structural information regarding the maturation process.

Fig. 1. Structure and function of phage MCP.

Phage MCP structure is highly conserved and provides three functions necessary for the formation of robust closed capsids. (A) The essential functions of MCP are shown in their role in capsid assembly. Self-assembly: Soluble MCP + SP subunits assemble into higher-ordered structures. QE: Sequence-identical MCP monomers can adopt different conformations, which can assemble into different higher-order units, in particular hexons and pentons. QE allows for the assembly of larger capsids incorporating more subunits than a simple T = 1 icosahedron, but can also result in nonviable products like sheets, tubes, and spiral shells. Maturation: Correct assembly results in a meta-stable enclosed prohead. After removal of the SP, MCP undergoes a coordinated transformation into a highly stable mature head. (B) Ribbon diagrams of the HK97 fold that provides these functions. The MCP of HK97 itself (PDB 1ohg) is shown above, and phi29’s MCP gp8 is shown below. (C) The four canonical domains and individual secondary structural elements (SSEs) are labeled on a schematic of gp8. The C-terminal domain has been removed for clarity. Features are named according to type, domain, and a number going from most to least conserved (across species and QE or maturation conformations). Features that are folded differently in some of our observed structures are colored gold, while those that do not change are colored light blue.

Bacillus subtilis bacteriophage phi29, the subject of this study, has a prolate capsid with C5 symmetry and 47 unique quasi-equivalent positions (not counting the variation necessary to accommodate the portal at the unique vertex) (13–15). In contrast, a T = 4 isometric phi29 capsid has the same number of MCP copies (240), but only four unique quasi-equivalent positions. There are also two regions with distinct geometries and markedly different curvatures in the prolate C5 capsids: hemi-icosahedral endcaps with T = 3 QE, the smallest observed T number in phage (therefore with the tightest curvature), and a cylindrical central region (with zero net curvature along the prolate axis). The local environments experienced by MCP in these regions are therefore less equivalent than in larger capsids, where changes in curvature are distributed more smoothly over the many capsomers. The mechanisms by which HK97-like MCP capsomers adopt different curvatures have not been described in detail. It is possible that different mechanisms are used in flatter versus more tightly curved regions. Phi29’s geometry also provides another unusual opportunity for studying capsid assembly. Isometric phages have roughly spherical proheads; each MCP subunit is thus equidistant from the center such that attached SPs have equal access to oligomerization partners regardless of the quasi-equivalent position in the capsid. This is not true of prolate capsids; therefore, the interactions between phi29 MCP and SP may also have differences that can inform the general case. Perhaps due to these differences, no state analogous to the smaller, bumpier, less angular isometric procapsids common to other phages has been observed in phi29, where procapsids appear to be the same size as mature capsids. However, since DNA packaging seems to be the trigger for MCP transformation and capsid expansion in most phages, it is possible that phi29 capsids may transform as well, but in a more subtle ways, potentially involving an intermediate conformation.

Here, we report near-atomic resolution cryo-TEM image reconstructions of phi29 capsids before the DNA-packaging motor is bound and immediately after DNA packaging has completed. From these, we found three notable structural features not seen in any other phage: (i) concave capsomers in the equatorial region of the capsid; (ii) open gaps around the pentameric vertices of empty proheads; and (iii) resolvable density of complete SP, including higher-order oligomers. The concave capsomers result from changes in the A-domain and allow a greater range of curvature in the capsid. The open gaps around pentons are the result of an MCP conformation halfway between the prohead and mature states and provide a path for N-arm migration to the capsid exterior during capsid maturation. The scaffold is present at two distinct binding sites on MCP and can assemble as different higher-order oligomers. These features extend our understanding of phage MCP and SP function/s and indicate how macromolecular complexes coordinate their functions more generally.

RESULTS

Cryo-TEM reconstructions of phi29 procapsids before and after DNA packaging

For this analysis, we reexamined cryo-TEM data from four samples initially imaged for different projects, two investigating the mechano-chemistry of the phi29 DNA packaging motor (16) and two exploring whether fusing small RNAs to the pRNA (the phage-encoded RNA that attaches the DNA packaging motor to the prohead) could be used as a general method to reconstruct small RNAs via cryo-TEM. Micrographs from the former project samples contain a mixture of empty and filled particles, and thus, reconstructions from the same micrographs allowed size comparison of the two states, excluding the possibility of magnification errors. Unlike other phages with distinct immature and mature capsids, the density for the phi29 capsid in both states superimposed nearly perfectly other than a slight protrusion of pentons in fully packaged particles. This indicates that the internal pressure resulting from DNA packaging does not cause mechanical strain-based expansion (fig. S1). The highest-resolution map from the motor project was of filled particles and reached a resolution of 3.0 Å as measured by Fourier shell correlation cutoff of 0.143 (FSC 0.143) (fig. S2A). This map was used to build the 47 unique subunits of the C5 asymmetric unit. The highest-resolution map for an empty capsid came from the second sample of the pRNA project; these particles are incapable of packaging DNA, and thus, maturation could not have been triggered. This map also reached a resolution of 3.0 Å (FSC 0.143) (fig. S2B), and the atomic coordinates built into the asymmetric unit of the DNA-filled particles were adjusted to fit the density of the MCP asymmetric unit of the empty capsid. Subparticle images of pentons were extracted from particle images of whole capsids from both pRNA samples and then used for focused refinement of pentons with clear SP density, as discussed below. Density for the flexible SP was found only in maps of empty capsids. This SP density was of lower resolution, but secondary structure was clearly visible and allowed unambiguous fitting of the x-ray crystal structure [Protein Data Bank (PDB) 1noh] (17). Imaging conditions and reconstruction statistics are summarized in table S1. All representative density maps show clear continuous density for the backbone and most side chains of the MCP (Fig. 2). Atomic model validation statistics are listed for the representative models and maps in table S2.

Fig. 2. Cryo-TEM structure of pre- and post-DNA filled phi29 proheads.

(A) Electron density maps of phi29 are shown for both empty (bottom) and DNA-filled (top) states. The interior is shown in cutaway on the left, while the capsid exterior is shown in the center. (B) Extracted density of the structural units that make up the capsid are shown, a hexon above and an isolated monomer below. (C) Fit of atomic models into density is shown for the 3.0-Å resolution full capsid map (top) and 3.7-Å empty capsid map (bottom).

Phi29’s small prolate capsid reveals new mechanisms of QE

All tailed phage capsids are formed from different types of capsomers, and different sizes and shapes of capsid can require more or fewer types of capsomers. T = 3 and T = 7 icosahedral capsids have two kinds of capsomers, pentons and hexons, and each copy of both exists in identical local environments. In T = 3 capsids, each hexon abuts three pentons, whereas in T = 7 capsids, only one penton contacts each hexamer. Capsids between T = 9 and T = 13 contain two types of hexon, one of which does not contact any pentons (18). Phi29’s prolate capsid has four different types of capsomers with corresponding local environments, shown as P1-P3 and H1-H6 in Fig. 3B. All pentons contact five hexons, but hexons exist in three different environments. “True” icosahedral hexons (H1, H6) occur in the endcaps of the capsid and have the same local environment as a hexon in a T = 3 isometric icosahedron, i.e., contacting three pentons and three other hexons. “Semi” icosahedral hexons (H2, H5) are at the edges of the endcaps, and one of its neighboring pentons is replaced by an “equatorial” hexon. Equatorial hexons (H3, H4) occur in the cylindrical middle region and only contact one penton. H6 has several localized distortions that accommodate its contacts with the portal and other features at the unique vertex and has been described as a separate type of capsomer in other works (15, 19). For the purposes of a general discussion on QE, we treat H6 and H1 as equally true icosahedral hexons.

Fig. 3. Forty-seven quasi-equivalent positions are fulfilled by only four basic conformational classes.

The structural variability of phi29 MCP (gene product 8; gp8) in post-DNA packaged capsids is shown by aligning all 47 individual monomers in the C5 asymmetric unit. Monomers are shown after being aligned to their least variable residues including βA1 (A) and in their original position in the capsid (B). Four distinct structural classes were identified, which corresponded to specific local environments at quasi-equivalent positions in the asymmetric unit. Monomers are colored according to their class in all panels. Eleven Pentameric monomers are colored in pink; 16 Even and 16 Odd icosahedral hexon monomers are colored light blue and green, respectively; four monomers in the Equatorial conformation are dark blue. The four canonical structural domains are shown separately at the bottom, each viewed from a direction that best shows its variability.

We characterized the symmetry of these four capsomer types by measuring root mean square deviation (RMSD) values of capsomers with rotated copies of themselves, as summarized in table S3. Pentons have nearly perfect fivefold symmetry. Note that unlike isometric icosahedra where fivefold symmetry is imposed at all pentons, in these prolate maps fivefold symmetry was only imposed on the penton along the C5 symmetry axis. Both icosahedral hexons are closer to threefold symmetry than sixfold, especially the true icosahedral hexons, consistent with their occupation of positions analogous to threefold axes in isometric T = 3 icosahedra. The equatorial hexons, however, only had strong twofold symmetry. Twofold symmetric hexons are also present in T = 4, 7, or 9 icosahedra, where hexons can occupy icosahedral twofold symmetry axes. However, phi29’s equatorial hexons are distinct in that they are concave rather than convex; no concave capsomers have been described previously. The different symmetries of these capsomers require different numbers of quasi-equivalent monomers; fivefold symmetry in pentons allows only one monomer conformation, threefold symmetry in true and semi-icosahedral hexamers allows two different conformations, while twofold symmetric equatorial hexamers must have three. Global C5 symmetry of a T_end = 3 T_mid = 5 capsid lattice allows up to 47 MCP conformations (235/5, the number of subunits in the C5 asymmetric unit). However, the local environments of the capsomers reduce this number. We superimposed all 47 individual monomers in the asymmetric unit and found only four general classes of conformation (Fig. 3): a Penton conformation with 11 copies per asymmetric unit, an Equatorial conformation with 4 copies, and “Even” and “Odd” conformations, both with 16 copies each. The Penton and Equatorial classes are quite distinct, but the Even and Odd classes are similar to one another. Most of the structural variance among the 47 monomers in the asymmetric unit depends on the overall curvature of the domains that make up the capsomer edges. This edge is curved like a parentheses symbol [“(“] in Pentameric monomers, has reduced curvature in Odd and Even icosahedral monomers, and is nearly straight in Equatorial monomers. These changes in curvature are accompanied by a rotation of the monomer’s P-domain relative to the plane of the capsomer. We did not observe any differences in inter-residue bonding associated with differences in the bending state, indicating that the discrete bending states observed are not directly encoded in phi29’s peptide sequence, but result from the local environments created by the quaternary structure of the capsid.

In all published phage capsid structures, different shapes of capsomers are accompanied by differently folded A-domains. Phi29’s convex equatorial hexons include two monomers with the Equatorial conformation, which includes a differently folded A-domain (Fig. 3). The first helical turn in Equatorial αA1 uncoils, and its residues extend a loop that, in all other conformations, is well ordered and binds to αA2. Density corresponding to the unbound loop is blurry in cryo-TEM density maps, indicating conformational flexibility. When this loop does not bind to αA2, αA2 is able to shift toward αA1, resulting in a more acute angle in the A-domain, which binds αA1 of the adjacent monomer at a different angle. This different binding angle in the concave hexons allows for zero net curvature in the cylindrical equatorial region along the prolate axis.

Variation in capsomer shape provides a range of intra-capsomer curvature, and variation in the angle at which capsomers meet provides a range of inter-capsomer curvature. These capsomer angles are mostly determined by the structure at the pseudo-twofold capsomer edge junctions, where the conformational displacements are the greatest (Fig. 4). In the prohead state of phage capsids, these edge junctions consist mainly of direct contact along the length of P-domains parallel to the edge. In the mature state, however, there is no direct contact between opposing P-domains, and the junction between them is composed of the αP1 and N-arm of two other monomers that are oblique to the edge, as shown in Fig. 4A (bottom). This arrangement positions the inter-capsomer binding sites at the ends of long, flexible regions, where they can be displaced considerably by minor changes in bond angle over a large number of residues. The relatively large displacements of αP1, the N-arm, and the E-loop can be seen in Fig. 3. We chose representative residues to define planes in each capsomer that could be used to calculate the inter-capsomer angles (Fig. 5). Two clusters of inter-capsomer angles resulted: around 42° for most inter-hexon junctions, and around 20° for junctions involving pentons, or between an equatorial hexon and its semi-icosahedral hexon neighbor along the prolate axis (Fig. 5, bottom row). Repeating this analysis for HK97 found different inter-capsomer angles, but these also fell within a ~20° range (fig. S3). This range of angles is larger and more discrete than is suggested by the range of deflections in the monomers themselves (Fig. 3A). Furthermore, different hexon/hexon junctions with angles of 42° and 20° are composed of the same Even and Odd monomers, with only their arrangement reversed. This is possible because the two regions that bind are displaced in opposite directions when the outer edge of the monomer flexes. The high-curvature junctions are the result of two inward-displaced regions binding together, and the low-curvature junctions result from two outward-displaced regions binding together, as illustrated in Fig. 4. However, the low-curvature junctions are still not flat enough for the geometry of the cylindrical equatorial region, and this shortfall of inter-capsomer flexibility is what necessitates concave hexons.

Fig. 4. Arrangement of different monomer conformations amplifies differences and allows a greater range of angles at the twofold capsomer edge junctions.

Even conformations are shown as blue, and Odd are shown as green. (A) The interactions that bind capsomers at their edges are shown for both prohead (top) and mature (bottom) conformations. During maturation, the bowed P-domains extend out past their inter-capsomer neighbors into the junction. The mature form involves four monomers per junction, while the prohead form only involves two. (B) Two cross-sections of a monomer are important for the mature junction, one along the outside edge of the capsomer and the other perpendicular. The curvatures at these cross-sections are opposite; the Odd conformation has greater curvature parallel to the capsomer edge junction, while the Even conformation has greater curvature in the perpendicular direction. (C) These opposite curvatures allow junctions with different arrangements of the same conformations to form different angles. Degree of local curvature is exaggerated in (B).

Fig. 5. Both inter- and intra-capsomer bending is required to create a small prolate capsid.

Characteristic capsomers that span the full range of local symmetries and intra-capsomer angles are highlighted in a capsid map in the top left panel. Residues chosen to define planes for each capsomer are shown as colored spheres: Asp¹⁰³ as blue, Lys¹¹⁸ as yellow, and Leu²²⁶ as red. The cylindrical equatorial region requires an overall angle of 0° between H2 and H4, but the minimum inter-capsomer angle present in the capsid is 20°. The twofold symmetric concave equatorial hexon H4 has an intra-capsomer angle of −20° that resolves this problem. Hexons in the endcaps are planar due to the quasi-equivalent requirements of a T = 3 lattice.

Pre-packaging pentons have a conformation intermediate between prohead and mature head

Phi29 pre-packaging empty capsid differs from its DNA-packaged full capsid in the pentons, which have an intermediate MCP conformation with similarities to both mature and prohead conformations of MCP seen in other phage (Fig. 6). The N-arms in mature-state MCP help make up the inter-capsomer edge junctions and extend to the capsid exterior. In phi29, they form extensive contacts wrapped around an in-capsomer neighbor. Intermediate-state penton N-arms remain in the capsid interior, resulting in a gap at the edge junctions with the five neighboring hexons. A 12-Å pore is also open in the center of these pentons, similar to those seen in the hexons of P22 and SPP1 proheads (10, 20). The intermediate-state N-arm is folded into an α helix that forms a coiled coil along with the “spine” helix, resulting in greater curvature of the spine helix and thus the P-domain, which can also be observed in the proheads of P22, HK97, 80α, T7, and T5 (10, 21–24). The angle between the A- and P-domains of phi29’s MCP in pentons is different before and after DNA packaging; similar changes in this angle can be observed before and after prohead expansion in HK97, P22, and T7 (10, 21, 22). On the other hand, the curvature of the P-domain in phi29 pre-packaging pentons is more similar to that of the mature state in other phages. Thus, the MCP conformation of the pentons in phi29 pre-packaging capsids likely represents a previously unknown intermediate state between the canonical prohead and mature states. We propose that this intermediate conformation results from prohead monomers being “stretched” by the expansion of their surrounding hexons, yet prevented from refolding into the mature conformation by interactions of the interiorly positioned penton N-arms with SP. In a previously reported reconstruction of empty phi29 capsids (15), the MCP density for the interior N-arm was weaker and interpreted as SP. This interpretation is understandable because the density for the N-arm in those maps is present in two locations simultaneously, presumably due to particles with both conformations contributing to the reconstruction. Since protein can only be in one place at a time, the weaker N-arm density corresponding to the interior conformation was assumed to be the scaffold, which superficially resembles the N-arm in its largely helical content. In our empty map, the exterior density is not present, while the interior density is more clear and can be unambiguously assigned to the N-arm rather than SP.

Fig. 6. Pentons in the unfilled prohead have an intermediate conformation.

MCP in these pentons has characteristics that can be observed in proheads of other phage (helical N-arm, rotation of P-domain relative to the A-domain), as well as characteristics of mature conformations (low curvature in the P-domain). This intermediate state is likely due to the competing influences of the SP, which stabilizes a prohead-like conformation, and the neighboring expanded hexons, which stabilize a mature head-like conformation. (A) Phi29 capsids likely first assemble in an as-yet unobserved true prohead form similar to proheads of other phage (top left). The hexons expand first and pull the corner junctions with them, which results in a penton that is expanded at the periphery, but retains a prohead-like “puckered” core (top center). This creates gaps at the penton/hexon edges. When SP has dissociated from the pentons, MCP expands to a fully mature conformation, which closes the gaps (top right). (B) Density maps of intermediate (orange) and mature (magenta) pentons are highlighted, with surrounding hexons in teal. (C) Atomic models of intermediate (orange) and mature (magenta) MCP are shown as ribbon diagrams of the whole molecule (except the C-terminal BIG2 domain) and stick ensembles of each domain isolated and rotated to show variation.

SP oligomers localize at immature pentons

Low-pass filtering reconstructions of the pre-packaging capsids show a five-armed structure of density associated with each penton in the capsid interior that cannot be accounted for by MCP (Fig. 7A). Hence, we suspected that this density might correspond to the SP. A crystal structure of the phi29 SP has been determined, with a structure resembling two arms of the five-armed density (PDB 1NOH) (17). In both, the arms have an angle of about 72° between them and terminate in an “arrowhead” structure of a four-helix bundle. The SP tetramer crystal structure can be fitted as a rigid body into two of the five arms with no additional structural adjustments, indicating that the asymmetric unit of the crystal structure captured a biologically relevant intermediate. Further, in the crystal structure, the binding between the two dimers is such that a third dimer could be similarly attached in series (Fig. 7D). This attachment can be repeated to create a five-armed decamer that resembles the observed density (Fig. 7E). SP appears to be at least partially unstructured in solution studies and dynamically associates as dimers and other oligomers (17, 25–27). This dynamic nature is believed to be important to SP function (25). Subparticle maps of intermediate pentons refined with C5 symmetry had density matching the five-armed structure described previously. In these reconstructions, in addition to the dimers, density extends from the central blob to a cleft on the MCP N-arm (Fig. 7B). A scrap of high-resolution density, which cannot be accounted for by the MCP sequence, is present at this cleft and appears bound there (Fig. 7, F and G). The residues of the helix-turn-helix motif in the SP, which has been shown to control binding to MCP in P22 (28, 29), can be fitted into the fragment of high-resolution density near the N-arm (Fig. 8). This fit suggests that interactions between D17 on the scaffold and R68 on the MCP likely play an important role. A salt bridge between these two residues may weaken as DNA and associated counterions fill the capsid, reducing the Debye length via electrostatic screening. The location of the MCP-binding region at the N-terminal end of phi29’s SP is notably in contrast to P22 (28) and 80alpha (23), as well as phage like HK97 and T4, in which the C-terminal domain (CTD) of the scaffold is linked directly to the N-terminal domain (NTD) of the MCP as a single polypeptide.

Fig. 7. Visualization of scaffold oligomers at pentons.

(A) Low-pass–filtered reconstructions of pre-packaging phi29 capsids show density under the pentons. (B) Extraction of the pentons as subparticles and refinement with C5 symmetry showed densities very similar to the crystal structure of phi29’s scaffold protein. (C) The scaffold protein associates into dimers with a coiled-coil arrowhead shape. The N terminus is at the end of the short helix, and the C terminus is at the end of the long helix. The “helix-turn-helix” motif connecting the arrowhead to the tail has been shown to bind MCP in other phage. (D) Superimposing the top dimer from one model of the scaffolding tetramer with the bottom dimer of another tetramer gives a hexamer. (E) Continuing this process gives a spiral decamer, which resembles the experimental density when viewed axially. (F) Sorting the penton subparticles generates classes with two, one, or zero (not shown) fully dimerized SP pairs. The remaining three, four, or five dimers in the decamer are partially unzipped at the N-terminal end to allow binding to a cleft in the MCP N-arm, shown as small yellow densities. In sites with well-defined full-dimer density, density for the N-arm binding region of that dimer is weak or not present. Low-occupancy scaffold-binding sites on MCP N-arms, which expose the SP-binding cleft, are indicated by arrows. (G) An additional class from a later sample shows three dimers, in different locations from those seen in (F). The central blob is an artifact of ensemble and symmetry averaging, and it is not visible at the higher contour levels used for (F) and (G).

Fig. 8. Interactions between scaffold and MCP.

(A) SP (orange) binds to the N-arm of maturation intermediate-state MCP (magenta) at two distinct binding sites. (B) Monomeric SP, likely the helix-turn-helix region near the N-terminal region (NTR), binds to a cleft in the N-arm. (C) Dimeric SP binds its NTR to the distal face of the MCP N-arm. This binding site is likely sterically blocked in the true prohead state in which hexons have not expanded. (D) A mockup of our proposed true prohead state, generated by fitting an intermediate-state penton into the density of a penton-adjacent hexon, suggests that partially dimerized SP could connect two capsomers across the edge junctions.

SP binding to the N-arm cleft requires separation and partial unfolding of the SP arrowhead domain. Thus, monomers of the scaffolding dimer must at least partially dissociate such that the MCP-binding region of the SP can bind at the cleft in the MCP N-arm. The ability of SP dimers to partially dissociate may be promoted by “pinning” via strong cation-pi interactions between two HIS50 residues, visible as a bridge of density connecting the two helices near their midpoint (fig. S4). The relative energies of SP dimerization and partial dissociation of dimers to bind N-arm clefts must be balanced to synchronize the events of capsomer maturation. To understand how SP/MCP interactions might influence the as-yet unobserved “true prohead” state of phi29, we created a model of immature/immature capsomer junctions by replacing a mature hexon that contacts an immature penton with another immature penton (Fig. 8D). While we would have preferred to replace the mature hexons with immature hexons, there are no atomic structures of a putative immature hexons, never mind hexons with visible scaffold protein. Nonetheless, we believe that using immature pentons accurately captures the crucial MCP/SP interactions at the edge junctions circumscribing the pentons. In this model, a scaffold dimer can bind N-arms on separate capsomers across the edge junction while still remaining mostly dimerized. SP residues 20 to 31 in this model are unstructured and have no interactions with other peptides, consistent with previous findings that these residues alone have increased H/D exchange when SP is incorporated into a prohead (30). The PISA (Proteins, Interfaces, Structures, and Assemblies)–estimated buried surface area (31) of this model is very close to that of an isolated SP dimer (Table S4). This small difference in buried surface area, and thus binding energy, suggests a dynamic equilibrium between scaffold/MCP binding and full SP dimer formation. In higher-order oligomers of scaffold protein, the 72° angle between dimers matches the angle and spacing of N-arms in adjacent monomers within pentons and may therefore serve to keep SP near its MCP-binding site even after it has detached, increasing the chance it can rebind before the N-arm unfolds and migrates to the capsid exterior. The overall effect of this arrangement would be to stabilize pentons in the immature state. However, if one monomer in a dimer is separated from its N-arm, as must happen when hexons expand and their N-arms migrate to the capsid exterior, unbinding of the second becomes favored.

Symmetry expansion and classification separated the ensemble-averaged structure in Fig. 7B into several constituent classes. In most classes, one or more dimers of SP are well resolved, with the other SP only visible as the monomeric MCP-bound fragment. The most populated (30% of pentons) class shows a single dimer of SP, with its NTDs positioned between two MCP N-arms (Fig. 6E, left). The cleft of the N-arm, which would bind to this dimer, is unoccupied, supporting that full dimerization is mutually exclusive with binding MCP. This result also suggests that, while only one SP NTD is bound to the N-arm, its CTD is still associated in a dimer, and thus, at least 10 molecules of scaffold protein are present at each penton. The unbound SP NTD from the other monomer in each dimer was likely previously bound to the N-arm of an MCP monomer in a neighboring hexon, as seen in SPP1 prohead intermediates and as indicated in Fig. 8D (20). No five-armed densities resulted from C1 reconstructions; the structures seen in whole-capsid reconstructions are likely the result of averaging randomly oriented dimers and tetramers. Density for the arm pointed toward the equatorial region of the prolate capsid is the weakest, suggesting that dimers are least likely to form in this position.

We observed an alternate positioning of scaffold oligomers in one preparation of pRNA proheads (Fig. 7G). In these, a dimer with well-defined density is positioned in contact with one N-arm of the MCP, rather than floating between two (Fig. 7F). The density for these dimers is of higher resolution than the free-floating NTDs, with bulky side chains identifiable at the interface. The C-terminal ends of three dimers are resolved as tubular densities in a helical stack (Fig. 7G), consistent with the higher-order oligomerization structure we extrapolated from the tetramer crystal structure (Fig. 7D). However, the helical axis of this stack is rotated approximately 45° away from the C5 axis of the penton. This misalignment with the penton may be a necessary intermediate step in breaking up the stable association with pentons so SP can leave the capsid when its function is fulfilled. In our model of the immature unexpanded prohead capsomer junctions, these dimer-binding sites on N-arms are sterically blocked by the N-arms of neighboring capsomers. Dimer binding would thus only become possible once hexons had expanded. The shorter N-arms of published isometric phage proheads do not have any structures analogous to this dimer-binding site.

Unique conformation at the PE-shoulder stabilizes gap-opening intermediate state

The penton MCP at the gap at inter-capsomer junctions displays a unique structural feature at the junction between the P-domain and E-loop, which likely stabilizes the gap. The sheet βE1 (Fig. 1C) partially unzips to form the intermediate state—four pairs of residues compared to eight in the mature state—with the outer strand swapping to join βP2 (fig. S5). As a result, the PE-shoulder, which connects the N-arm to the “body” of the rest of the MCP, is longer and more flexible. In published structures of other phage, the PE-shoulder and βE1 have the same conformation in both prohead and mature states, and there is no open gap at the capsomer junctions. The PE-shoulder is buried in both these states, but exposed to solvent in our phi29 intermediate state; thus, its ability to refold may stabilize the intermediate conformation and the capsomer junction gap long enough to be observed by cryo-TEM. During maturation of all known phage proheads, the N-arm H-bonds with βE1 after it has disassociated from the scaffold. This may serve to destabilize the capsomer junction gap and close the expanded capsid by extending βE1. If so, the loop connecting βP1 and βP2 to βE1 serves as a “switch,” allowing controlled opening and closing of the capsomer junction gaps. Since the scaffold-mediated assembly and maturation processes of tailed phage proheads are highly conserved, this switching mechanism at the phi29 PE-shoulder likely occurs in other phage systems, but with a lifetime too short to be observed. Evidence that gaps occur during maturation at the junctions around pentons includes the “wiffle ball” capsids with missing pentons described for P22 and HK97 (32–34), which we have also observed in some expression systems of phi29 proheads (fig. S6). Intermediate pentons are bound to the rest of the capsid much less securely, at threefold junctions alone. PISA analysis estimates that removing one would expose 8100 Ų of buried surface, compared to 24,000 Ų for a mature penton. The βP2/βE1 connecting loop is the only structural feature that is not embellished in any published HK97-like MCP structure, which further suggests the PE-shoulder switching mechanism may be common to other phage.

DISCUSSION

The intermediate state of phi29 pentons and their associated scaffold oligomers shows that the interaction of the N-arm helix with the spine helix is sufficient to lock MCP in a partially immature conformation, even in a local environment with neighboring mature conformations. Being blocked from fully adopting either stable state results in an open gap at the capsomer edge junction that is not seen in any fully immature (procapsid) or mature phage structures. The N terminus of the phi29 MCP in the intermediate state is close to this edge junction gap, allowing it to easily migrate to the capsid exterior. The PE-shoulder and βE1 in intermediate pentons have a conformation that has not been observed before. We hypothesize that this PE-shoulder conformation stabilizes this intermediate state and keeps the edge junction gap open long enough for the N-arm to migrate through this gap from the interior to the exterior of the capsid. SP forms higher-order oligomers of dimers at intermediate-state pentons. Scaffold/MCP binding mutually stabilizes both the scaffold oligomerization and the prohead/intermediate state of the MCP. The interaction with SP is strongly affected by the local symmetry of MCP: The 72° angle between SP dimers in the matches that of pentons, but not hexons.

Based on our results, we propose the following maturation pathway (Fig. 9). Prohead assembly begins with MCP bound to scaffold in the monomer binding site (Fig. 9A). Dimerization of MCP-bound SP brings two capsomers together to accelerate shell formation (Fig. 9B). Once a capsomer is fully surrounded by other capsomers, its SP dimers begin to form transient tetramers (Fig. 9C). At the pentons, these tetramers are more stable due to the common 72° angle between dimers in the tetramer and between MCP monomers in a penton. This results in the formation of higher-order oligomers with a symmetry axis coincident with the fivefold axis of pentons (Fig. 7E). Oligomerizing dimers are thus pulled away from MCP in neighboring hexons, promoting unbinding of SP from the MCP N-arms (Figs. 7, B and F, and 9D). Particles as shown in Fig. 9 (C and D) correspond to the unexpanded prohead states that have been observed in other phages (35). At hexon/hexon junctions, SP tetramer formation alone may be enough to disrupt scaffold/MCP binding. The unbound N-arm helices uncoil, which destabilizes the prohead conformation of the MCP. Once a hexon has been stripped of enough scaffold, it stochastically transitions to an intermediate state analogous to that described above for pentons. This opens a gap through which the hexon’s N-arms can migrate to the capsid exterior (fig. S5C). The N-arm then forms an intermolecular β sheet with the E-loop of an in-capsomer neighboring monomer, which stabilizes the mature conformation. Maturation is accelerated in neighboring hexons, until only pentons with their stabilized scaffold oligomers remain in the intermediate state (Figs. 7 and 9E). With the N-arms of hexons having migrated out from the capsid interior, the SP dimer-binding sites are exposed at pentons (Figs. 8 and 9F). The scaffold migrates to these dimer-binding sites, which breaks the symmetry match that stabilized the pentons in their intermediate state (Figs. 7G and 9G). The SP then dissociates completely from MCP (Fig. 9H), and pentons are thus free to mature (Fig. 9I). This model provides a mechanism for expansion to begin only when a hexon is in contact with three pentons that are themselves in contact with five hexons, something that only occurs at the endcap opposite the portal vertex. Assuming that assembly begins with the portal, as has been reported (36), the opposite endcap will only be finished when the capsid is fully closed, and thus, our model would provide an intrinsic mechanism for delaying expansion until assembly of a closed shell has completed.

Fig. 9. Proposed tailed phage capsid assembly pathway.

(A) Soluble MCP monomers bound at their N-arms to SP monomers. (B) Dimerization of SP brings together bound MCP, accelerating prohead assembly. (C) Dimer-dimer binding of MCP-bound SP is most stable at pentons, shown as purple, inducing strain that weakens SP-hexon binding. (D) When SP disassociates from hexons, this strain pulls it away from the binding site. (E) Unbound hexameric MCP expands, opening gaps at the capsomer edge junctions (black arrows). N-arms migrate to the capsid exterior. (F) Dimerization of SP NTRs competes with the N-arm cleft for binding monomers. (G) SP dimers bind at the newly exposed dimer-binding site, increasing the strain on remaining cleft-bound SP and making an (F) to (E) transition less likely. (H) Oligomerization of SP leaves most penton N-arms unbound. (I) Penton N-arms migrate to exterior as MCP expands, removing all binding sites for SP. Capsomer edge junction gaps close (black arrows). The size of red arrows between lettered states indicates estimated forward and reverse rates for the associated state transition.

Our proposed pathway for tailed phage capsid maturation suggests several targets for mutation experiments to test its validity and universality. For example, phage T4 encodes separate MCP for hexons and pentons, both of which require cleavage of the N-terminal residues for maturation. Changing the site of this cleavage for the hexon MCP only should result in stable particles with intermediate folding states at the pentons. Mutations that stabilize βE1 would be expected to block the PE-shoulder refolding that stabilizes the intermediate state and should therefore slow or block the maturation process. Conversely, mutations that block formation of the intermolecular E/N β sheet may slow or block maturation beyond the intermediate state.

Our results also solidify some general principles of quasi-equivalence in HK97-like phage MCP. First is that refolding of the A-domain controls the symmetry of capsomers that can be assembled. Thus, edges of the A-domain are a “control factor” in which minor amino acid changes can effect major changes in capsid geometry. Single-point mutations in the A-domains of P22 and lambda MCPs have been found to produce long cylindrical “polyheads” even in the presence of SP (37, 38), consistent with an inability of the new A-domains to conform into pentons. Second, the presence of only four conformational classes in the 47 unique positions in phi29’s capsid and two clusters of inter-capsomer angles shows that QE is neither a deterministic decompression of information from primary sequence to secondary, tertiary, and quaternary structure of the completed capsid nor a passive accommodation by MCP to overall capsid geometry. Emergent two-way interactions between the capsomer units control the tertiary structure of monomers in those units. These two-way interactions suggests that the order in which capsomers expand determines the final conformation reached; thus, phi29’s long-lived penton intermediate state may be critical to ensure that the maturing capsid reaches the global energy minima.

Phi29 is an unusual phage, but its unique features allowed us to capture elusive intermediate states, some of which are likely generalizable to all phage and thus provide a more complete understanding of the capsid maturation process. More broadly, study of seemingly specialist details can often reveal fundamental principles that would otherwise escape notice.

METHODS

Particle production, grid preparation, data collection, and initial image processing

As described in Results, data for this study were initially collected for two projects that are unrelated to capsid assembly. The unexpected presence of scaffolding density in these reconstructions motivated the analysis presented here. For particles that were originally analyzed in regard to DNA packaging, particle purification, grid preparation and freezing, data collection, initial single-particle analysis and three-dimensional reconstruction, model building, and validation protocols were described previously (16).

The purification protocol for particles originally analyzed as part of the pRNA-related project was the same as that used for the DNA packaging project and referenced above. Subsequent grid preparation/freezing, data collection, initial image processing, and model building will be described in detail elsewhere. Briefly, samples were incubated on Quantifoil (R2/1), 200-mesh holey carbon grids for 40 s before being blotted for 17 s using a Leica EM GP2 automatic plunge freezer. After blotting, samples were vitrified in liquid ethane. Grids were then transferred to the Titan Krios microscope equipped with a Gatan K3 camera and located at the University of Texas Medical Branch (UTMB) Sealy Center for Structural Biology for data collection. The accumulated electron dose for each movie stack was ~40 e/Ų, and the pixel size was 1.089 Å at the detector. Subsequent contrast transfer function (CTF) corrections indicated a defocus range of ~1 to 3 μm. Motion correction, CTF correction, particle picking, and initial image processing were carried out using cryoSPARC. Final resolutions are presented in table S2.

Focused refinement

Penton re-boxing and extraction from partially signal-subtracted images (39) was carried out using the “local reconstruction” plugin for scipion (40). Using the Chimera tool “Segger,” we made two masks in which one of the pentons from the upper and lower ring of five was removed. The penton opposite the motor was not included in this analysis. Subparticles were refined with C5 symmetry and then classified based on the presence of SP. Scaffold-containing subparticles were refined again and then were symmetry-expanded to C1 using relion_particle_symmetry_expand. The expanded set was classified without further alignment to identify different arrangements of the SP.

Atomic model building and analysis

An initial atomic model of the MCP was built in Coot using the density map of a single monomer extracted by Segger from the full-size map from the first session. This model was duplicated and fitted into extracted density corresponding to the C5 asymmetric unit of the capsid using Chimera. This initial model was refined using the Phenix real-space refinement protocol. The model was further refined into later higher-resolution maps and into the DNA-filled map after manual rebuilding. Atomic coordinates for the SP were taken from published crystal structure PDB 1NOH (17) and refined into focused refinement densities using PHENIX, with secondary structure restraints at the less-resolved regions. Buried surface areas were estimated using the Proteins, Interfaces, Structures, and Assemblies (PISA) interactive tool at the European Bioinformatics Institute website (31).

Supplementary Materials

This PDF file includes:

Tables S1 to S4

Figs. S1 to S6