|Home | About | Journals | Submit | Contact Us | Français|
Bacteriophage P4 is dependent on structural proteins supplied by a helper phage, P2, to assemble infectious virions. Bacteriophage P2 normally forms an icosahedral capsid with T=7 symmetry from the gpN capsid protein, the gpO scaffolding protein and the gpQ portal protein. In the presence of P4, however, the same structural proteins are assembled into a smaller capsid with T=4 symmetry. This size determination is effected by the P4-encoded protein Sid, which forms an external scaffold around the small P4 procapsids. Size responsiveness (sir) mutants in gpN fail to assemble small capsids even in the presence of Sid. We have produced large and small procapsids by co-expression of gpN with gpO and Sid, respectively, and applied cryo-electron microscopy and three-dimensional reconstruction methods to visualize these procapsids. gpN has an HK97-like fold and interacts with Sid in an exposed loop where the sir mutations are clustered. The T=7 lattice of P2 has dextro handedness, unlike the laevo lattices of other phages with this fold observed so far.
Bacteriophage P2 is a temperate, double-stranded (ds) DNA phage of the Myoviridae family that infects Escherichia coli and other enterobacteria (Bertani and Bertani, 1970; Bertani and Six, 1988; Nilsson and Haggård-Ljungquist, 2006). P2-like prophages are common in the environment (Breitbart et al., 2002), are present in about 30% of strains in the E. coli reference collection (Nilsson et al., 2004), and likely play an important role in horizontal gene transfer in bacterial evolution.
The P2 virion consists of a 60 nm diameter isometric capsid, or head, attached to a 135 nm long contractile tail tipped by a base plate with six tail fibers attached. The capsids consist of 415 copies of capsid protein, gene product (gp) of the N gene (gpN) arranged with T=7 icosahedral symmetry (Dokland et al., 1992). P2 capsids are assembled as roughly spherical precursors called procapsids from gpN (40.2 kDa), several copies of a gpO scaffolding protein and a dodecameric connector or portal (gpQ) that forms a unique vertex through which the DNA is subsequently packaged (Fig. 1A) (Lengyel et al., 1973; Chang et al., 2008). At some point during the assembly process, gpN is processed by the removal of 31 N-terminal residues, yielding a mature cleavage product called N* (36.7 kDa) (Lengyel et al., 1973; Rishovd and Lindqvist, 1992). Likewise, gpQ is processed to Q* by the removal of 26 N-terminal amino acids, while gpO is cleaved between residues 141 and 142, leaving an N-terminal fragment, O*, that remains inside the mature capsid (Rishovd et al., 1994; Chang et al., 2008). These cleavage events are thought to be carried out by gpO, which contains a serine protease domain and possesses autoproteolytic activity (Wang et al., 2006; Chang et al., 2009). The proteolytic activity of gpO resides in the N-terminal half of the protein, while the C-terminal 90 amino acids comprise a scaffolding domain, which is required and sufficient for promoting capsid assembly (Wang et al., 2006; Chang et al., 2008; Chang et al., 2009).
Circular P2 genomic DNA substrates (33.6 kb) are packaged into these empty procapsids through the action of the terminase proteins gpM and gpP (Bowden and Modrich, 1985), upon which the capsid undergoes major structural transitions that lead to a more angular mature capsid (Lengyel et al., 1973; Marvik et al., 1994a; Chang et al., 2008) (Fig. 1B). The tails, which are assembled via a separate pathway, are attached to the gpQ connector at the unique vertex (Lengyel et al., 1974). A head completion protein, gpL, is added to the final assembly product (Chang et al., 2008).
P4 is an 11.6 kb replicon that can exist either as a plasmid or integrated into the host genome like a prophage (Briani et al., 2001; Deho and Ghisotti, 2006). It is genetically unrelated to P2, except in the cohesive (cos) ends. P4 does not normally get mobilized at high frequency and lacks genes encoding major structural proteins. In the presence of P2, however, P4 gets packaged into phage particles using structural proteins supplied by the P2 helper (Six, 1975). The capsid produced in the presence of P4 is smaller (45 nm) than the normal P2 capsid and contains 235 copies of gpN arranged with T=4 symmetry (Dokland et al., 1992; Lindqvist et al., 1993). This size difference is effected by the P4-encoded Sid (Size determination) protein (Barrett et al., 1976), which forms an external scaffold surrounding the P4 procapsids (Marvik et al., 1995) (Fig. 1C). P4 sid mutants fail to form small capsids; however, P4 DNA can still get packaged as dimers or trimers into large capsids at low efficiency (Shore et al., 1978). Mutants in gpN, called sir (sid responsiveness) render the capsid protein unable to form small capsids even in the presence of functional Sid (Six et al., 1991). Furthermore, certain mutations in Sid, named super-sid or nms (N mutation sensitive), can act as second-site suppressors of the sir mutations and enable the formation of small capsids even on a P2 Nsir background (Kim et al., 2001). Capsid assembly is completed by removal of Sid, cleavage of gpN, gpO and gpQ, and addition of the Psu decoration protein (Dokland et al., 1992; Rishovd and Lindqvist, 1992; Dokland et al., 1993; Chang et al., 2008) (Fig. 1D–F).
We showed previously that P4-size procapsids can be produced by co-expression of gpN and Sid alone (Dokland et al., 2002). Likewise, P2 procapsids can be produced by co-expression of gpN with a protease–defective version of gpO generated either by truncating the protein at the N-terminus or by mutating one of the three active site residues (Wang et al., 2006; Chang et al., 2009). In this paper, we present three-dimensional reconstructions of P2 and P4 procapsids produced by co-expression of gpN with gpO and Sid, respectively. The gpN capsid protein has the HK97 fold seen in all other tailed dsDNA phages studied to date (Bamford et al., 2005; Johnson and Chiu, 2007). Unusually, the P2 procapsid has a T=7 dextro lattice, unlike the T=7 laevo lattices of all other bacteriophage capsids with the HK97 fold observed to date. gpO is not ordered with icosahedral symmetry and does not appear in the reconstruction. In the P4 procapsids, Sid forms a dodecahedral cage that surrounds the procapsid and interacts with the gpN hexamers. Sid–gpN interactions occur at a loop in the apical domain of gpN where the sir mutations are located.
P4 procapsids were produced by the co-expression of gpN and Sid, while P2 procapsids were produced by co-expression of gpN and protease-deficient forms of gpO, either with or without the gpQ portal protein (Dokland et al., 2002; Wang et al., 2006; Chang et al., 2008) (Table 1). The procapsids were purified on sucrose gradients and prepared for cryo-EM as previously described (Dokland et al., 2002; Huiskonen et al., 2004; Dokland and Ng, 2006). The gpQ portals were not incorporated into capsids, but remained on top of the sucrose gradients after purification. The resulting shells were indistinguishable from those formed in the absence of gpQ.
P4 procapsids are hollow, somewhat angular shells with an average diameter of 40 nm (Fig. 2A). These particles have a characteristic “hairy” appearance resulting from the external Sid scaffold (Wang et al., 2000; Dokland et al., 2002), and lack an inner core, since they were produced in the absence of internal scaffolding. Very few aberrant particles were seen. P2 procapsids are roughly spherical shells with an average diameter of 54 nm and a shell thickness of about 4–5 nm (Fig. 2B). This appearance is quite distinct from the angular, thin-walled shells typical of expanded capsids (Chang et al., 2008). The procapsids appear to be partially filled, presumably by the gpO internal scaffolding protein. A few aberrant, incomplete or broken shells were also present.
Reconstructions of P4 procapsids produced both by in vitro assembly from purified gpN and Sid protein and from the co-expression of gpN and Sid were previously described (Wang et al., 2000; Dokland et al., 2002). The best resolution attained in those studies was 21 Å. The P4 procapsid reconstruction presented here (Fig. 3A) was generated from 2286 particle images and reaches a resolution of 9.5 Å based on the 0.5 Fourier shell correlation (FSC) criterion. At this resolution, several features corresponding to ordered α-helices can be distinguished, forming the basis for protein modeling based on the HK97 gp5 structure, as described below.
In the P4 procapsid, gpN is arranged into twelve pentamers and 30 elongated hexamers that sit on the icosahedral two-fold symmetry axes (Fig. 3A). The T=4 icosahedral symmetry dictates that there are four non-equivalent gpN subunits in the shell, denoted A–D, which are organized into an A5 pentamer and a (BCD)2 twofold symmetric hexamer (Dokland et al., 1992). The capsomers consist of protruding domains clustered around the symmetry axes, and basal domains that interact via two types of similar, trivalent contacts (Fig. 3A): a strict threefold contact at the icosahedral threefold axis (type 1), and a quasi-threefold contact between two hexamers and a pentamer (type 2).
The external scaffold made of Sid forms a dodecahedral cage that makes trivalent connections at the icosahedral threefold axes and interacts with the gpN hexamers at the icosahedral twofolds (Fig. 3A). This interaction involves primarily the B subunits.
The P2 procapsid was reconstructed from 7582 particle images to a resolution of 9.9 Å based on the 0.5 FSC criterion. The P2 procapsid reconstruction (Fig. 3B) shows the characteristic T=7 structure of capsid protein clustered in twelve protruding pentamers and 60 elongated (“skewed”) hexamers, similar to those seen in other bacteriophage procapsids, including lambda, HK97 and P22 (Dokland and Murialdo, 1993; Conway et al., 1995; Jiang et al., 2003). Unusually, the T=7 lattice in P2 is right-handed (dextro), rather than the laevo lattice found in other bacteriophages. (Although handedness cannot be determined from projection images, the correct hand of the reconstruction can be ascertained by comparing the detailed structure of the subunits to that of bacteriophage HK97—see below.) There are thus seven non-equivalent gpN subunits in the P2 shell, organized into an A5 pentamer and a BCDEFG hexamer (Dokland et al., 1992). Contacts between capsomers in the P2 lattice are similar to those in P4, except that in addition to the type 1 and type 2 contacts mentioned above, the T=7 P2 lattice has an additional trivalent contact (type 3) formed between three hexamers that connect near the twofold symmetry axis (Fig. 3B).
We showed previously that gpO binds to gpN and is present in procapsid shells in about 120–160 copies per procapsid (Wang et al., 2006). P2 virions contain about 70 copies of O* (Geisselsoder et al., 1982). In spite of this, there is no evidence of the gpO internal scaffold in the reconstruction, apart from a diffuse inner density only slightly higher than the background (not shown). Presumably, gpO is not ordered symmetrically in the capsid.
All tailed dsDNA bacteriophage (order Caudovirales) structures observed to date share a common structural motif (Bamford et al., 2005; Johnson and Chiu, 2007), referred to as the “HK97 fold” after the first bacteriophage structure solved at high resolution, the gp5 capsid protein of HK97 (Wikoff et al., 2000; Helgstrand et al., 2003; Gertsman et al., 2009). We therefore expected the same motif to be present in gpN. The HK97 fold has a characteristic two-domain structure, with a P domain that includes a long α helix called the “spine helix” and a three-stranded β sheet, and an A domain of mixed α/β structure that represents an insertion into the P domain (Fig. 4A,B). The A and P domains are separated by a flexible “β hinge” that is a source of conformational diversity in many systems (Gertsman et al., 2009; Teschke and Parent, 2010). Other features include the E loop, a long β hairpin motif that protrudes from the P domain, and an extended N-terminal arm that contains little secondary structure.
In spite of the absence of sequence similarity between gpN and HK97 gp5 (or any other bacteriophage capsid protein), secondary structure prediction revealed a similar distribution of structural elements in gpN and gp5 when the sequences were aligned using the program FUGUE (Shi et al., 2001) (Fig. 4A). (The first 104 amino acids of gp5 constitute a “delta” domain, which is cleaved during capsid assembly, and was absent from the gp5 crystal structure. It was omitted from the alignment.) The N-terminal 55 residues of gpN have two long predicted α-helices (residues 2–18 and 35–55) that do not correspond directly to the extended N-terminal conformation in the gp5 crystal structure. However, bacteriophages ε15, P22 and 80α also have α-helices in the N-arm (Jiang et al., 2008; Parent et al., 2010; Spilman et al., 2011). The gpN-gp5 alignment also predicts a number of insertions in the gpN sequence; the longest one (24 residues) was located between helices α3 and α4 (Fig. 4A).
The HK97 gp5 procapsid structure (PDB ID: 3E8K) was placed into the electron density for the P2 procapsid. There was good correspondence between the A and P domains of gp5 and the protruding and basal parts of the hexamer density (Fig. 4C), in an orientation similar to that of the HK97 procapsid (Gertsman et al., 2009). The spine helix (α3) of gp5 was immediately recognizable as a tubular density in the map. Although the A domain density was generally less well defined than the P domain, perhaps due to flexibility, tubular densities corresponding to helices α5 and α6 were easily recognizable (Figs. 4C). However, after fitting the P domain of gp5 using the spine helix as a landmark, the A domain fell outside the density. The angle between the two domains was therefore adjusted at the β hinge to allow both domains to fit into the density (Fig. 4C), similar to the adjustments used to allow for the conformational variation in HK97 and P22 (Gertsman et al., 2009; Teschke and Parent, 2010). Fitting gp5 into the P4 procapsid reconstruction gave the same result (not shown).
It was clear that the structure could not be made to fit reasonably into a map with the opposite hand (Fig. 4D), showing that the dextro hand of the T=7 lattice was correct. To further validate this result, we carried out simultaneous classification against two HK97-based models, one with a T=7 laevo and one with a T=7 dextro lattice. Only the T=7 dextro lattice resulted in a reconstruction.
Based on the fitting and the secondary structure alignment (Fig. 4), we were confident that HK97 gp5 constituted a suitable model for gpN. Using the gpN-gp5 alignment as a guide, we generated a model for gpN with the I-TASSER threading server, specifying HK97 gp5 (PDB IDs 1OHG or 3E8K) as a template (Zhang, 2008). The top resulting I-TASSER models all had the gp5 fold and differed primarily in the N-arm, the E loop (residues 70–104) and the insertion loops (Fig. 5A).
In order to fit the gpN I-TASSER model into the P2 and P4 procapsid reconstructions, it was first split at the β hinge that divides the A and P domains (Fig. 5B). The A and P domains were fitted individually into the maps as rigid bodies, using clearly recognizable structural motifs as a guide, such as the α3 spine helix and the β sheet (βD, βH and βJ) in the P domain and helices α5 and α6 in the A domain (Fig. 5B–D). This was followed by minor additional relative adjustments of motifs within the domains. Some of the interspersing loops, which could not be modeled with confidence, were omitted from the model. Part of the E loop, for which there was no clear density, was omitted. The individual strands of the three-stranded β sheet could not be resolved in the map. However, they match a flat density that is connected to α2 at one end of the P domain (Fig. 5C, D). There were only small differences in gpN between P2 and P4, which could be accommodated by small shifts in the A domain helices (Fig. 5C,D).
Helix α5 in the A domain is part of a motif (residues 192–224) that includes four of the five known sir loci (Table 2; Figs 4A and and5B)5B) (Six et al., 1991). Hence, we refer to this motif as the “sir loop”. Two of the sir loci reside in α5 (A217 and L221), while two are at the apex of the loop (D206, Y207). A fifth sir locus, M184, is located in the P domain, just before the sir loop. The sir loop matches an arch-like density in the A domain (Figs. 5C, D). This density makes contact with the Sid external scaffold in the P4 procapsid (Fig. 5D), consistent with a role of the sir mutations in affecting Sid–gpN interactions (see below).
The I-TASSER model predicts two α-helices (α0 and α1) in the N-terminal arm of gpN (Figs. 4A, ,5A).5A). In contrast, the gp5 N-arm in the HK97 virion contains little secondary structure and extends across to the E-loop of the adjacent subunit (Helgstrand et al., 2003) (Figs. 4A,B). In the HK97 procapsid, however, the N-arm points towards the inside of the shell (Gertsman et al., 2009) and in bacteriophages ε15, P22 and 80α, there is a helix in this position (Jiang et al., 2008; Parent et al., 2010; Teschke and Parent, 2010; Spilman et al., 2011). In the P2 and P4 maps, helix α1 most likely corresponds to a tubular density that extends from the A-P domain interface towards the interior of the shell, similar to α1 in P22 and 80α (Fig. 5B,C,D). An additional tubular density underneath the P domain most likely corresponds to α0, although the connectivity to the rest of the subunit is unclear (Fig. 5B,C,D).
Intra-capsomeric interactions in both P2 and P4 involve contacts between the E loop and the P domain of the adjacent subunit, and between α5 (the sir loop) and α6 of the adjacent subunit (Figs. 6A-D), similar to interactions observed in other phages, including HK97 (Helgstrand et al., 2003). Conformational variation between subunits is evident in the α6/α6' loop, which adjusts to fill in the central hole in the capsomers, and in the sir loop (Figs. 6A,B). The hexamer structures of P2 and P4 are remarkably similar, and almost perfectly superimposable (Fig. 6E), even though the P2 hexamers are not constrained by symmetry, while the P4 hexamers sit on icosahedral twofold axes (Fig. 3). The only obvious difference in a P4–P2 hexamer difference map is the external scaffold made of Sid (Fig. 6E).
Interactions between capsomers occur at trivalent contacts (Fig. 3), in which the P domains of adjacent subunits form a threefold annulus (Fig. 6F). These interactions are formed at the site of an insertion in gpN between βH and βJ (Fig. 4A), equivalent to the “P loop” of phages 80α and P22 (Parent et al., 2010; Spilman et al., 2011).
The external scaffold made of Sid consists of multiple segments of tubular density that connect in an annulus at the threefold axis and form an interlocking hook above the hexamers at the twofold axes (Figs. 3A and 6C–E). The size and shape of the tubular densities is consistent with α-helices. There is no homology model available for Sid; however, sequence analysis and biophysical characterization suggested that Sid is virtually entirely α-helical with a high propensity for coiled coil formation (Wang et al., 2000).
The scaffold spans a distance of about 9 nm from the twofold to threefold axes, constituting one icosahedral asymmetric unit. If Sid were 100% α-helical, it would have a length of 36.6 nm when fully extended. The tubular densities observed in the map account for a distance of about 25 nm—not taking into account connecting loops—which is thus roughly consistent with a single copy of Sid. The exact connectivity between the helices cannot be resolved in the map, and it is therefore impossible to say whether one Sid subunit stretches from twofold to threefold or if it has a more intertwined structure, e.g. stretching from threefold to threefold as a dimer. The super-sid (nms) mutations are clustered near the C-terminus of Sid (Kim et al., 2001). If we assume that these mutations are located at the gpN–Sid interface where the sir mutations are found, the C-terminus of Sid is most likely located in the finger-like protrusion that extends from the scaffold toward the sir loop (Figs. 5B,D and and6D6D).
Over the past decade it has become clear that all tailed dsDNA bacteriophages (order Caudovirales) are structurally related and share a common capsid protein fold (Bamford et al., 2005; Johnson and Chiu, 2007), first described at high resolution in the E. coli bacteriophage HK97 (Wikoff et al., 2000). Apart from the HK97 gp5 capsid protein, examples of this fold are found in the structures of T4, T7, ϕ29, P22 and ε15 as well as herpesviruses and the archaeal virus-like particle PfV (Baker et al., 2005; Fokine et al., 2005; Morais et al., 2005; Agirrezabala et al., 2007; Akita et al., 2007; Jiang et al., 2008; Parent et al., 2010). Thus, gpN was expected to conform to the same structure, in spite of a lack of detectable sequence homology. While the two folds are indeed similar, gpN differs from gp5 by having several insertions into its sequence, primarily in the P domain, conformational differences in the A domain, a partly disordered E loop and an ordered, α-helical N-arm (Figs. 4A and 5A,B). However, unlike e.g. P22 or T4 (Fokine et al., 2005; Parent et al., 2010), there is no additional, independent domain inserted into the gpN sequence.
One unexpected difference was the T=7 dextro lattice of the P2 capsid. Other T=7 phages observed so far, including HK97, P22, T7 and ε15, all have T=7 laevo lattices. (The papovaviruses have pseudo-T=7 dextro lattices, but this is based on an all-pentamer architecture (Belnap et al., 1996).) It is conceivable that this is an artifact of the expression system and that native procapsids have T=7 laevo hand, but this seems unlikely. We also considered the possibility that the sample contained a mixture of both lattices, but refinement against two models of opposite hands yielded only a T=7 dextro structure. In spite of this difference in handedness, however, interactions between subunits and capsomers are retained: gpN forms the same kind of trivalent contacts between P domains as in many other bacteriophages (Fig. 3B, ,6F).6F). This anomaly also illustrates the point that protein chirality does not necessarily determine the handedness of the supramolecular assembly it forms (Chothia, 1991).
Non-equivalence in icosahedral capsids is reflected in conformational variability and differences in protein-protein interactions (Caspar, 1980; Dokland, 2000). In the P2 and P4 reconstructions, conformational variation is manifested through small shifts in the A domain helices, α5 and α6/α6' (Figs. 6A–D). However, there is no distinctive conformational difference that distinguishes P2 and P4. Rather, small perturbations in the capsid protein structure and interactions propagated over the whole lattice is sufficient to accommodate the difference in curvature between the T=4 and T=7 shells.
The P2/P4 system is unique in that two distinct capsid sizes can be assembled from the same capsid protein under control of the two scaffolding proteins, gpO and Sid (Marvik et al., 1994b) (Fig. 1). gpO promotes the formation of T=7 shells primarily by curtailing the intrinsic flexibility of gpN, much like a molecular chaperone (Dokland et al., 1992; Dokland et al., 1999; Dokland, 2000; Chang et al., 2009). In the presence of both gpO and Sid, Sid takes precedence, leading to the formation of T=4 capsids (Wang et al., 2006). Consistent with this observation, gpO interacts only transiently with gpN, while Sid is stably associated with the procapsids until expansion occurs (Wang et al., 2003; Chang et al., 2008).
Sid is unique in that it forms an external, ordered scaffold, unlike the internal scaffolding cores found in most dsDNA bacteriophages (Dokland, 1999). The other example of an external scaffolding protein is that of the Microviridae family of ssDNA viruses (Dokland et al., 1997), which is a globular protein, unlike Sid. Sid has a propensity for coiled coil formation and spontaneously forms trimers and higher oligomers in solution (Wang et al., 2000). Sequence prediction suggests a high likelihood of disorder in solution. Many proteins are known to possess this kind of “intrinsic disorder” and become ordered only upon binding to a substrate (Uversky and Dunker, 2010). Once bound to gpN hexamers, the elongated Sid molecules constitute a tether that would constrain the possible shell assemblies to the smaller T=4 shell. This oligomerization with a stably bound Sid–gpN complex thereby overrides the transient, chaperone-like action of gpO.
We previously suggested that the Sid scaffold was made up of 120 copies of Sid protein, based on measurements on Coomassie-stained gels, and proposed an arrangement in which two copies of elongated Sid subunits extend in tandem from the twofold to threefold axes of symmetry (Dokland et al., 2002). However, the P4 procapsid reconstruction shown here is consistent with the presence of only 60 copies of Sid, most likely in an arrangement where one copy of Sid stretches from the twofold to the threefold axes (Fig. 6C)
In P2 and P4, as in many dsDNA bacteriophages, capsid maturation involves cleavage of the capsid protein. Cleavage occurs after assembly (Marvik et al., 1994a), and pre-cleaved protein does not assemble correctly (Wang et al., 2000). Cleavage occurs prior to DNA packaging, however, and in procapsids isolated from a wildtype infection, gpN is already processed to its mature N* form (Rishovd and Lindqvist, 1992; Marvik et al., 1994a; Chang et al., 2008). In the P2 procapsid particles used here, cleavage of gpN and gpO does not occur because a protease-defective gpO was used (Wang et al., 2000; Chang et al., 2009). The 31-residue N-terminal peptide that is cleaved off from gpN includes the predicted helix α 0 (Fig. 4A). By analogy with other phage capsid proteins, especially P22 (Parent et al., 2010), it is likely that a tubular density underneath the P domain corresponds to the α0 helix (Fig. 5C,D). This helix might contact the gpO scaffolding protein, in which case cleavage could lead to scaffold release. Alternatively, cleavage might destabilize intersubunit interactions in preparation for expansion. Indeed, experimental removal of the N-terminal 29 residues by trypsin treatment was previously shown to cause spontaneous capsid expansion of P4 procapsid shells (Wang et al., 2003). The presence of gpO might play a role in delaying the expansion until DNA packaging occurs [the P4 procapsids that were expanded in vitro (Wang et al., 2003) did not contain gpO].
It was previously hypothesized that the sir mutations affected the flexibility in a presumed hinge region in gpN (Six et al., 1991; Kim et al., 2001). This hypothesis was based on the clustering of the sir mutations roughly in the middle of the gpN sequence and the observation of an apparent two-domain structure in the low resolution reconstructions of P2 and P4 mature capsids (Dokland et al., 1992). Variation in the observed angle between these two domains appeared to support this idea. This hypothesis was made before any high resolution structures of bacteriophage capsids had been solved. In light of the subsequent structures of HK97 gp5 and numerous other bacteriophage capsid proteins (Wikoff et al., 2000; Fokine et al., 2005; Jiang et al., 2008; Parent et al., 2010), it is now clear that this model is incorrect. Although the HK97 fold has two domains, their sequences are intertwined, with the A domain essentially forming an insertion in the P domain (Fig. 4A,B). Consequently, the sir mutations end up largely clustered in one loop (the sir loop) at the apex of the A domain (Fig. 5B). This loop forms the primary, if not the only, interaction point between Sid and gpN in the P4 procapsid (Fig. 5D). Thus, we conclude that the sir mutations most likely disrupt gpN–Sid interactions, either directly or indirectly, through changes in loop conformation. sir3, which affects M184, is located in the sequence preceding the sir loop, and might be more likely to affect loop conformation or flexibility.
The nms or “super-sid” mutations are clustered near the C-terminal end of Sid, and it was previously suggested that these mutations led to a strengthening of the Sid lattice at the threefold interaction point (Kim et al., 2001; Dokland et al., 2002). In light of the above results, it is more likely that the super-sid mutants suppress the sir phenotype by restoring interactions that were disrupted by the sir mutations. The C-terminus of Sid is therefore most likely located near the gpN–Sid interface, in the finger-like protrusions of the scaffold at these points (Fig. 5B,D).
The P2/P4 system represents a fascinating example of molecular piracy, in which a replicon usurps the structural gene products of an unrelated replicon for its own mobilization and spread. Satellite viruses are not uncommon in eukaryotes, but these viruses generally encode their own capsid proteins. Hepatitis delta virus, for example, uses envelope proteins supplied by hepatitis B virus, but encodes its own nucleocapsid protein (Sureau, 2006). P4 is different in that it is a complete and independent replicon with a highly specialized machinery for helper exploitation, depends on the helper for structural proteins, and has the ability to redirect the capsid assembly process. As it turns out, the P2/P4 system is not the only example of such a mechanism. In Staphylococcus aureus, genetic elements called S. aureus pathogenicity islands (SaPIs) are mobilized by specific helper phages (Lindsay et al., 1998; Novick et al., 2010) and are packaged into phage-like transducing particles using structural proteins supplied by the helper phage (Poliakov et al., 2008). In some cases, the SaPI capsids are smaller than those normally formed by the helper. In the S. aureus system, capsid size determination apparently depends on two internal scaffolding proteins (Dearborn et al., 2011; Spilman, 2011).
The plasmids used are listed in Table 1. pLucky7, which expresses gpN and Sid under control of separate T7 promoters from a chimeric vector, was previously described (Dokland et al., 2002). pET16b-derived plasmids pSW101 and pJRC49 express protease-deficient truncated and mutated versions of gpO, respectively, in tandem with gpN under control of the same T7 promoter (Wang et al., 2006; Chang et al., 2009). The Q gene was inserted after OΔ25 and N in pSW101 to generate plasmid pCMRP1 that co-expresses OΔ25, gpN and gpQ.
All cloning was performed in the E. coli recA- strain DH5α (Invitrogen). For expression, clones were transformed into E. coli strains BL21(DE3) or BLR(DE3) (Novagen). Expression was induced by IPTG and after 2 hr of growth, the cells were lysed in buffer containing 10 mM Tris-HCl pH 7.4, 200 mM NaCl, 1 mM PMSF, 1% Triton X-100, and 0.5% deoxycholate by freeze-thawing and high pressure disruption in a French Press (Thermo Fisher) or in an Emulsiflex EF-C3 high-pressure cell disruptor (Avestin Inc., Ottawa). Procapsids were harvested from the clarified lysate by pelleting at 110,000×g followed by one or two cycles of purification on 10–40% sucrose gradients in 50 mM Tris-HCl, pH 8.0, 100 mM or 1 M NaCl, 10 mM MgCl2 for 2 h at 110,000×g, 4 °C. Gradient fractions (1 ml each) were collected manually from the top of the gradient. Fractions containing procapsids were pooled, pelleted by centrifugation for 40 min at 178,000×g and resuspended in a low-ionic strength buffer containing 10mM Tris-HCl, pH 8.0, 20 mM NaCl and 2 mM MgCl2.
Cryo-EM was done by standard methods: the sample was applied to either non-glow discharged, chloroform-washed lacey grids (Electron Microscopy Sciences, Hartfield, PA) or glow discharged Quantifoil R2/2 grids (Quantifoil Micro Tools, Jena), blotted for 1–3 sec with filter paper, plunged into liquid ethane using a manual plunger, and transferred to a Gatan 626 cryo-sample holder (Baker et al., 1999; Huiskonen et al., 2004; Dokland and Ng, 2006). All samples were observed in an FEI Tecnai F20 field-emission gun (FEG) electron microscope operated at 200 kV. Images to be used for reconstruction were collected on Kodak SO-163 film under low dose conditions, at magnifications of 50,000x or 62,000x and typical defocus values of 1–3 μm.
The P4 procapsid images were digitized on a CreoScitex EverSmart scanner with a step size of 4.545 μm (220 dots/mm) and a 3x binning factor, leading to a pixel size of 2.73 Å/pixel with respect to the sample. P2 procapsid images were digitized with a Nikon CoolScan 9000ED film scanner with a raster step size of 12.7μm (2000 dpi), corresponding to 2.05 or 2.54 Å/pixel.
Three-dimensional reconstruction was done with EMAN versions 1.8-1.9 (Ludtke et al., 1999; Ludtke et al., 2001; Ludtke et al., 2004) and with AUTO3DEM (Yan et al., 2007). The contrast transfer function (CTF) curves were determined interactively using the program CTFIT in the EMAN suite (Ludtke et al., 1999; Saad et al., 2001) and a phase correction applied to the images. For P4, the existing low resolution reconstruction was used as a starting model, while for P2, the STARTICOS routine was used to generate a model from scratch. To check for handedness, the images were simultaneously classified against two models created by placing the HK97-based model in T=7 lattices of opposite hands. The final maps from EMAN were used as starting models for AUTO3DEM refinement. The final maps were calculated to 8Å resolution, and an inverse B factor of 200–250 Å2 was applied for map sharpening using bfilter from the Bsoft suite (Heymann and Belnap, 2007). Map visualization, segmentation and modeling were done using UCSF Chimera (Pettersen et al., 2004; Goddard et al., 2007). The P2 and P4 procapsid maps were deposited in EMDB with accession codes EMD-25276 and EMD-25273, respectively.
Prediction of secondary structure was done with the PROF (Ouali and King, 2000) and GOR (Garnier et al., 1996) servers. The gpN sequence was aligned with the HK97 gp5 structure (PDB ID: 1OHG) using FUGUE (Shi et al., 2001), giving an initial alignment of the secondary structure elements of the two proteins. A gpN starting model was generated by I-TASSER (Zhang, 2008), specifying the HK97 gp5 structure (PDB IDs 1OHG or 3E8K; both resulted in essentially identical models) as a template. The model was separated into the A and P domains, which were fitted separately into the corresponding density in UCSF Chimera (Pettersen et al., 2004; Goddard et al., 2007). Minimal additional adjustments were made in the A and P domains, and some loops and other features that did not clearly correspond to density were omitted from the final model.
Funding for this work at UAB was partly provided by The National Institutes of Health grants R21 AI071982 and R01 AI083255 to T.D. This work was initiated while T.D. was at the Institute for Molecular Agrobiology, Singapore, and was funded by Singapore's Agency for Science, Technology and Research (A*STAR). We thank the Biocenter Finland National Cryo Electron Microscopy Unit, Institute of Biotechnology, Helsinki University, for providing services.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.