Transient contacts on the exterior of the HK97 procapsid that are essential for capsid assembly

Abstract

The G-loop is a 10-residue glycine-rich loop that protrudes from the surface of the mature bacteriophage HK97 capsid at the C-terminal end of the long backbone helix of major capsid protein subunits. The G-loop is essential for assembly, is conserved in related capsid and encapsulin proteins, and plays its role during HK97 capsid assembly by making crucial contacts between the hill-like hexamers and pentamers in precursor proheads. These contacts are not preserved in the flattened capsomers of the mature capsid. Aspartate 231 in each of the ~400 G-loops interacts with Lysine 178 of the E-loop (Extended Loop) of a subunit on an adjacent capsomer. Mutations disrupting this interaction prevented correct assembly, and in some cases induced abnormal assembly into tubes, or small, incomplete capsids. Assembly remained defective when D231 and K178 were replaced with larger charged residues or when their positions were exchanged. Second-site suppressors of lethal mutants containing substitution D231L replaced the ionic interaction with new interactions between neutral or hydrophobic residues of about the same size: D231L/K178V, D231L/K178I, and D231L/K178N. We conclude that it is not the charge, but the size and shape of the side chains of residues 178 and 231 that are important. These two residues control the geometry of contacts between the E-loop and the G-loop, which apparently must be precisely spaced and oriented for correct assembly to occur. We present a model for how the G-loop could control HK97 assembly and identify G-loop-like protrusions in other capsid proteins that may play analogous roles.

Introduction

The atomic structure of the T=7 icosahedral capsid of bacteriophage HK97, Head II (PDB ID: 1OHG),^{1; 2} revealed an unusual 10 residue glycine-rich loop in the subunits of the HK97 major capsid protein. This compact loop protrudes from the capsid surface at the C-terminal end of the backbone α-helix (Figure 1 (a)) of the major capsid protein subunit, and it appears to be tightly connected to the α-helices that it is sandwiched between (Figure 1 (b,c)). We named this peculiar feature the G-loop for the four Glycine residues in its amino acid sequence, GDGTGDNLEG. In this work, we set out to investigate what role, if any, the G-loop might play during HK97 capsid assembly.

Figure 1. The G-loop of the HK97 and encapsulin subunits.

(a). Subunit F of HK97 Head II (PDB ID: 1OHG) as viewed from the outside with the G-loop, E-loop and other elements color-coded and featured residues highlighted. (b). Zoomed view of HK97 G-loop in space-filled mode showing its compact form, the tight packing of the first and last glycines, and how neighboring helical residues anchor the loop into the subunit. G-loop residues are in orange and glycines in red. (c). the same view as in e, but with space-filled mode off. (d). Sequence logo for 135 HK97-like G-loop sequences shows the enrichment and depletion of residue types by position, including the strict conservation of the first and last G-loop glycines. (e). A subunit from the Thermotoga encapsulin (PDB ID: 3DKT) with G-loop and E-loop color-coded as in a. (f). Sequence logo for 97 3DKT-like Type 1 encapsulin sequences (T=1 microcompartments related to the Thermotoga encapsulin) showing the frequency of occurrence.

Normal HK97 capsid assembly (Figure 2) requires three capsid proteins, the portal protein gp3, the maturation protease gp4, and the major capsid protein gp5 ^{3; 4}. The portal protein is optional for plasmid-directed capsid production. The major capsid protein (or mcp) is made as a 42 kDa precursor, which when expressed by itself, produces Prohead I, a partly hollow T=7 shell made of 420 copies of gp5 arranged as twelve pentons and sixty hexons⁵. When the maturation protease is present, the mcp and protease co-assemble, with the protease inside the shell^{3; 4; 6}. Assembly activates the protease, which digests the ~100 residue N-terminal domain from each capsid subunit, leaving Prohead II, a hollow shell composed of cleaved major capsid protein (gp5*, 31 kDa).^{3; 5} DNA packaging apparently triggers the final transformation in which the particle expands and takes on the angular shape of Head II^{5; 7; 8}.

Figure 2. HK97 assembly pathway.

Hexamers and pentamers of the HK97 major capsid protein, gp5, co-assemble with the protease, gp4, to form Prohead I. After assembly, the protease cleaves the N-termini of each gp5, cleaves itself and the pieces exit, leaving Prohead II. Prohead II matures to form Head II via conformational changes that include expanding in size in a process that is normally triggered by DNA packaging. The portal protein is required to make a phage, but HK97 capsids can form without a portal, so it was not included for the experiments reported here. Capsid images courtesy of J. F. Conway, University of Pittsburgh. ALT: The assembly of the T=7 shell of HK97 normally begins with the formation of hexamers and pentamers of the major capsid protein that assemble into Prohead I, the earliest large precursor. Prohead I is converted by proteolysis to Prohead II before the final maturation transformation in which the round-shaped Prohead II expands and rearranges its subunits to become the angular Head II (Figure 2). Note that the conversion of Prohead I to Prohead II requires the incorporation of the maturation protease into the interior of Prohead I during the assembly phase.

In the Head II atomic model (PDB ID: 1OHG)^{1; 2}, which represents the mature conformation of the HK97 capsid, the G-loop makes no notable contacts; the nearest residue is ~4.5Å distant in the same subunit (Figure 1 (a))¹. However, in HK97 Prohead II⁹ (PDB ID: 3E8K), the stable precusor to Head II (Figure P), each G-loop makes an ionic contact with a subunit in a neighboring capsomer (Figure 3). These capsomers are the morphological building blocks of capsids. In HK97 there are 12 pentameric capsomers (one at each vertex) and 60 hexameric capsomers on the faces of the icosahedral capsid. The intercapsomer contacts, which occur between D231 on the G-loops and K178 on the E-loops of adjacent capsomers, are clearly present in the model of the procapsid, Prohead II, but not in the model of the mature capsid, Head II (Figure 3). The structures of the outer shells of Prohead I and Prohead II are very similar, so essentially identical D231-K178 interactions are assumed to be present in Prohead I, even though the model for Prohead I¹⁰ (a 5.2 Å structure, PDB ID: 3QPR) does not explicitly show the sidechains.

Figure 3. Interactions of D231 and K178 before and after expansion.

(a). A close-up view of a part of the Prohead II atomic model (PDB ID: 3E8K) showing the interactions of D231 and K178. (B). A close-up view of a part of the Head II atomic model (PDB ID: 1OHG) showing that D231 and K178 have moved far apart (18-20Å). Arrows point to the same pair of residues in the two structures. Both models shown have the truncated E-loop of the Prohead II variant used to solve the Prohead II structure and all subunits were truncated to start at residue 128 for simplicity and for making a movie of the same transition (shown in Supplemental Movie S1).

The hexons in Prohead II take the form of dome- or hill-shaped blisters, more akin to the pyramidal shape of the pentons in both prohead and head, while the hexons of the mature Head II are nearly flat. Because of this radical difference in topology, residues D231 in the G-loops are on the lower “hill sides” of all capsomers in Prohead II and are in position to interact with residues K178 on the E-loops of neighboring capsomers (Figure 3(a), and Supplemental Figure S1), while the D231 residues in Head II are not. The D231-K178 interactions vary a little in detail depending on the context in the Prohead II structure, with separation distances of 2.4-3.4 Å in the four types of dimer interfaces. In Head II, however, these inter-capsomer contacts are broken, and D231 and K178 are separated by 17-22 Å (Figure 3(b)) as a consequence of the radical conformational changes that take place during HK97 maturation (Figure 2, Figure 3 and Supplemental Movie S1). The loss of these contacts during maturation suggests that the contacts might be required for assembly, but not for the subsequent steps of maturation.

We set out to explore the G-loop's function and, in particular, to test whether the transient interactions between residues D231 and K178 had any role early in assembly that. By early in assembly, we mean during the formation of Prohead I, as outlined in Figure 2. We show that changing the residues present at positions 231 and 178 in the HK97 major capsid protein either blocks efficient assembly or changes the form into which the protein assembles. Our results show that the contacts between residues 231 and 178 play an important role in HK97 assembly and we present an induced-fit model of HK97 capsid assembly that appears to explain how this could occur.

The HK97 fold, is widespread in nature, so our results have implications for areas outside of phage capsid assembly. The HK97 fold has been detected in every tailed phage capsid that has been imaged at sufficient resolution, including P22^{11; 12; 13}, T4¹⁴, λ¹⁵, ϕ29¹⁶, T5^{17; 18}, P2/P4¹⁹, epsilon15²⁰, T7^{21; 22}, the tailed-phage-like archaevirus HSTV-1²³, in the base layer of the Herpesvirus (HSV1) major capsid protein²⁴ and in the bacterial comparments known as encapsulins²⁵. With renewed interest in phage therapy to combat multiple-drug-resistant bacterial pathogens^{26; 27; 28}, and an emerging interest in using bacteriophage capsids as platforms for nanotechnology²⁹ (http://www.nano.gov/) directed to the treatment of disease^{30; 31} and the creation of new technologies^{32; 33; 34; 35}, it is important to have a deep understanding of capsid structures and assembly pathways in order that they can be manipulated and utilized to the greatest extent.

Results

The G-loop is conserved in capsid proteins and encapsulins

Apparent G-loop homologs were easily detected in sequence alignments produced by BLAST³⁶ searches of GENBANK using the mature part of HK97 capsid protein as a probe. (See, for example the alignment displayed using this hyperlink: http://gigapan.com/gigapans/b806a9da80b4a66bcb1dd1b273d5b7a4/). These 10 residue G-loop homologs all 1) have a conserved relative location within the alignment, 2) start and end with Gly and 3) have one or two additional Gly residues at positions 3, 5, or 7, as shown in Figure 1(d). The space-filled representation in Figure 1(c) illustrates why the first and last glycine residues are absolutely conserved—not enough room is available for any larger side chains. Some of the BLAST hits were clearly identified as known phages, but many hits were from bacterial genome sequences and thus are likely to represent the capsid proteins of prophages or cryptic phages. A G-loop-like sequence was also detected in phage T5^{17; 18}, which shares the HK97 fold and in many sequences that align rather less well to HK97, such as those identified by searches using PSI-BLAST³⁷. Most of the G-loops detected are 10 residues in length, but some instances appear to have only 9 residues (see examples near the bottom of the alignment shown in the hyperlink cited above), as do the G-loops of the encapsulins. Encapsulins²⁵ are small icosahedral bacterial compartments that utilize the HK97 fold to make T=1 or T=3 shells. The Type I encapsulins, which make T=1 shells, have a G-loop ((Figure 1(e)) that is 9 residues long and highly conserved (Figure 1(f)) in sequence and with a conserved location within the protein sequence. The structural element in HK97 that contains the G-loop appears to include the few residues to the left and right of the actual loop that provide what appears to be a rigid attachment to the helices on both sides. This idea is supported by the finding of a limited repertoire of residue types at positions to the right and left of the loop in Figure 1(b); a similar situation is seen for the encapsulin G-loop in Figure 1(d).

Mutants which disrupt the ionic interactions between residues D231 and K178 disrupt assembly

An initial set of mutants were made in a plasmid that produces HK97 proheads upon induction^{3; 4}. The mutants included D231A, D231L, K178E, and K178L, and Δ-Loop, a deletion of nine loop residues that leaves only a single glycine (by changing GDGTGDNLEG to G). Expression from plasmids containing the mutants was induced under ³⁵S-Met labeling conditions and the soluble protein assemblies produced were compared to control plasmids using non-denaturing gels (Figure 4). Note that the relative mobilites of HK97 proheads and capsomers in agarose gels have been extensively characterized using purified samples.^{3; 6; 38; 39} None of the amino acid substitution mutants made any detectable prohead-like structures as shown in Figure 4(a) - proheads run as distinct bands in a native agarose gel^{3; 38; 39}, like the Prohead II and Prohead I bands seen in lanes 1 & 2 of Figure 4(a). However, each of the mutants made oligomers with all of the properties expected for unassembled capsomers³⁹. On agarose gels (Figure 4(a)), the mutant capsomers ran as diffuse bands that are more like spots than bands³⁹, similar to those seen for the wild-type capsomers present in the protease defective mutant (lane 2), but with appropriate changes in relative electrophoretic mobility. In non-denaturing polyacrylamide gels (Figure 4(b)), the capsomers were resolved into discrete hexamer and pentamer bands³⁹ with similarly altered mobilities that correlated with the altered charge of the mutations. The Δ-Loop mutant made only a reduced amount of a soluble structure that migrated similarly to capsomers in agarose gels (Figure 4(a) lane 6), but runs slower due to its loss of 3 negative charges. We repeated the expression experiments under conditions which produce higher concentrations of protein than in the radio-labelling experiments and confirmed that ample soluble capsomers, but no proheads, were made for this same panel of mutants (Figure 4c). Curiously, the D231A extract showed a new band that migrated slower than capsomers (Figure 4c, lane 4). We examined the D231A sample containing the extra band by electron microscopy (EM) and found small diameter, incomplete or irregular capsid-like particles (Figure 5 (a)) that appear to have initiated assembly, but which did not form complete shells. The material present in the extra band in the D231A sample was extracted from an agarose gel and visualized by EM (Figure 5 (b)), confirming that the extra band contained the small capsid-like particles and some very short tubes. The supernatants of the other mutants did not show any similar structures. The only other abnormal capsid protein assemblies observed for this set of mutants were a few very rare tubes in D231L (pellet) and K178L (pellet and supernatant, see example in Figure 5 (c)).

Figure 4. Mutants of D231 and K178.

Mutants at D231 and K178 produced no proheads but instead made capsomers which have electrophoretic mobility shifts that correspond to the predicted changes in their charge. Radiolabeled soluble proteins made from wild-type and mutant plasmids were displayed on (a), a 0.8% agarose gel and (b), a 7.5% native polyacrylamide gel. Proheads I and II run as sharp bands in the agarose gel, but do not enter the polyacrylamide gel. Capsomers (hexamers and pentamers) separate on the native polacrylamide gel, as indicated, but they run as fuzzy bands in agarose because they are poorly sieved by the gel. The predicted relative charge change for each sample is noted in the figure. (c). A 0.9% agarose gel showing assemblies of mutant capsid protein made in small scale cultures under high-level expression with pellet, supernatant and PEG precipitate fractions. The ladder of labels shows the migration positions of capsomers with predicted charge alterations.

Figure 5. Abnormal assembly directed by mutations at D231 and K178.

Fractions from small-scale expression experiments were visualized after negative staining with 1% uranyl acetate. Pellet fractions were mixed thoroughly and applied to grids without dilution, supernatants were first diluted 1/20 with 20 mM Tris-HCl/40 mM NaCl. (a). Abundant, small and poorly shaped partial capsid fragments seen in the supernatant fraction of mutant D231A. (b) Abnormal capsid fragments extracted from an agarose gel band of the supernatant fraction of mutant D231A. (c) An example of the rare wide tubes seen in the pellet of mutant K178L (also seen in the supernatant). (d) and (e) Small and poorly shaped partial capsid fragments seen in the supernatant fraction of mutant D231E. (f) Abundant tubes seen in the pellet of a K178D extract. (g) Abundant tubes and rare small particles (inset panels) found in a K178D/D231K supernatant fraction. The dark outlined inset panel labeled PII shows a wild-type Prohead II for comparison. (h) An example of the smaller diameter tubes seen in the pellet fraction of mutant K178D/D231K. (i) Wild type HK97 Prohead II. All images are presented at the same magnification.

The functional abilities of each of the mutants were also tested using plasmidphage complementation in which the cloned mutant genes on plasmids were challenged to supply HK97 major capsid protein to infecting phage lacking the ability to make the major capsid protein due to an amber mutation in gene 5. None of the substitutions at D231 or K178 was able to complement, as shown in Table 1.

Table 1. Spot complementation tests with amber mutants in gene.

5. Plasmid containing strains were tested for their ability to support the growth of phage with amber mutations in the major capsid protein gene (5). The values are all relative to the complementation achieved by wild-type versions of genes - these are marked with a“*”.

Form of gp5 present on test plasmid		Relative efficiency of plating (EOP)a when gene 5- amber phage was applied
wild type		1.0*
G-loop deletion	Δ-loop	<0.001
substitutions	D231L	<0.001
	D231A	<0.001
	K178E	<0.001
	K178L	<0.001
substitutions with larger residues of the same charge	D231E	0.0003
	K178R	0.0001
	D231E/K178R	0.0001
charge swap mutants and controls	D231K	0.0001
	K178D	0.0001
	D231K/K178D	0.001
pseudo-revertants	D231L/K178V	1.0
	D231L/K178I	1.0
	D231L/K178N	1.0

^aThe EOPs of gene 5^- amber phage on the strain expressing mutant proteins were measured relative to wild type as described in Materials and Methods.

Larger side-chain substitutions that preserve the charges of D231 and K178 fail to assemble

We changed both of these interacting residues to larger side-chains with the same charges to see if mutants that maintained the charge-charge interaction, but had larger sidechains could still assemble normally. Mutants D231E, K178R, and the double mutant D231E/K178R were made on plasmids and assayed for their phenotypes by complementation and by observing what structures were made after plasmid expression without radiolabeling. All three mutants were defective in complementation tests (Table 1), and the reason appears to be defective assembly, as shown below. All three of the new mutants (D231E, K178R, & D231E/K178R) produced soluble protein that was not assembled into stable structures larger than capsomers (Figure 6(a) lanes 1, 2, 3), except for mutant D231E, which produced an extra band of assembled protein which migrates slightly slower than capsomers (see “*” in Figure 6 (a) lane 4). When we examined the D231E supernatant sample by EM (Figure 5 (d) and (e)), we found that this mutant produced small, incomplete shells similar to those seen for mutant D213A (above) in addition to unassembled capsomers. Controls for this experiment included wild-type, a protease knockout and an unrelated capsid protein mutant (K103L) and all controls produced proheads (Figure 6(a) lanes 4, 5, 6). An active protease gene was present on all three mutant plasmids (D231E, K178R, & D231E/K178R), but no cleaved major capsid protein was observed (not shown), as expected since assembly into proheads is required for cleavage to occur ^{3; 4}. The presence of active protease on the plasmids ensures that any proheads that assemble will be cleaved and trapped in the Prohead II form. So wild-type HK97, and any mutants which assemble into proheads efficiently, all produce cleaved Prohead II, although a few mutants that assemble into proheads less efficiently produce a mixture of Proheads I & II ³⁸ (further discussion of this point follows later). Prohead II does not normally disassemble, although it is metastable in a different way, in that many treatments cause Prohead II to expand and convert to Heads³.

Figure 6. Effects of larger charged substitutions and exchanging the salt bridge residues D231 and K178.

Proteins from mutant and control plasmids were expressed using the mini-prohead protocol described under Materials and Methods. (a) The 0.9% native agarose gel shows that the iso-charge substitution mutants make unassembled capsomers, except for the partial capsids produced by mutant D231E marked by a “*” in lane 3. The proteins made by double-mutant D231K/K178D, single-mutant controls, and wild type were analyzed on (b) 0.9% agarose gels and (c) 12% SDS gels. Note that for K178D, the level of capsomers ((b) lane 5) is reduced relative to WT ((b) lane 2). Substantial amounts of gp5 protein in the K178D lysate was found in the pellet fraction in the form of tubes, as diagnosed by the presence of a high-molecular-weight ladder of uncleaved, but crosslinked major capsid protein, which is an indicator of the formation of abnormal, tube-like structures. The abnormal migration of D231 mutant proteins is discussing in the legend to Figure 4.

Effects of exchanging the charges of salt bridge residues D231 and K178

We reasoned that the positions of the two interacting charged residues might be interchangeable, so we made the charge-swap mutant D231K/K178D and the two single-mutant controls. We tested these new mutants for their ability to complement a gene 5 amber mutant phage and found that they were all defective (Table 1), so reversing the polarity of this salt bridge was not successful in producing functional capsids. The assembled protein complexes produced by these mutants after plasmid expression were compared to controls (Figure 6(b and c). G-loop mutant D231K produced capsomers (Figure 6(b)), including both hexamers and pentamers (not shown), which have lower mobility due to their altered charge. Lack of further assembly by mutant D231K is emphasized by the lack of cleavage of the major capsid protein (Figure 6(c)) in the presence of active protease. Mutant K178D, the other control for the charge swap, produces unassembled capsomers in reduced amounts compared to other mutants, because about half of the protein was found in abundant tangles of very long, narrow tubes of capsid protein (Figure 5(f)). The K178D tubes have about the same diameter (50 nm) as normal proheads (Figure 5(i)). The presence of the large structures was initially diagnosed by the material trapped in the well of the agarose gel (Figure 6(b) lane 4) and the ladder of cross-linked 42 kDa protein bands in the SDS gel (Figure 6(c) lane 6).

The charge-swap mutant D231K/K178D gave similar results to D231K in that much of the protein appeared in the agarose gel as capsomers , while no prohead band was observed (Figure 6(b) lane 5). Curiously the major capsid protein in mutant K178D/D231K was equally abundant in the pellet, supernatant and PEG fractions, suggesting that the bulk of this mutated protein had assembled into forms larger than capsomers. Abnormal assembly into larger structures was also indicated by the ladder of high-molecular-weight bands present in the pellet fraction (Figure 6(c) lane 9). Electron microscopy revealed that the K178D/D231K mutant protein does indeed assemble into a variety of structures (Figure 5(g-h)). The K178D/D231K pellet contained 30-40 nm diameter long straight tubes (Figure 5(g)), while the supernatant contained narrow long tubes (Figure 5(g)) and abundant masses of tangled bent tubes. The K178D/D231K mutant also made a small number of isolated particles that had a wide range of small diameters and regular or irregular shapes, all of which appear to be composed of capsid protein.

Suppressors of mutant D231L replace K178 with small uncharged residues

In order to discover if other residue pairs might sucessfully replace D231 and K178, we selected one very defective substitution, D231L, and used recombineering⁴⁰ to move the mutation from the plasmid into an HK97 lysogen with the goal of selecting second-site revertants. Since this mutation was defective in complementation tests (Table 1), we expected that few or no phage would be made upon induction of this lysogen, and this was the case. The HK97 lysogen carrying the gene 5 D231L mutation produced a yield of phage that was ~10⁹-fold less than produced by a wild-type lysogen induced in parallel. When we analyzed the few plaques that appeared after the induction of the HK97 gene 5 D231L lysogen, we found that they were all pseudo-revertants with second-site mutations, in that they still carried the original gene 5 D231L mutation, but they had acquired mutations at another position in the same gene that compensated for the original defect. All six of the secondary mutations analyzed were changes in codon 178 of gene 5 and changed lysine 178 to valine, isoleucine, or asparagine, as shown in Table 2. The recovery of functionally sufficient pseudorevertant double-mutant pairs show that interactions between the G-loops and adjacent E-loops are indeed important, but that an ionic interaction is not necessary.

Table 2. The second-site suppressors of mutant gp5 D231L in HK97 revertants.

The first column shows the relevant wild-type residues and codons in HK97 gene 5, the second shows the parent mutations, and the third shows the genotypes of phage present in seven randomly selected pseudo-revertant plaques.

WT residues & codons	parent mutation & codon	residues & codons in revertant
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178V (GTA)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178V (GTA)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178I (ATA)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178I (ATA)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178I (ATA)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178N (AAT)
D231 (GAT), K178 (AAA)	D231L (CTC)	D231L (CTC), K178N (AAC)

D231L pseudorevertants assemble less efficiently than wild-type

The region of HK97 gene 5 containing each pair of mutations was moved into our standard HK97 expression plasmid and also into a plasmid with a knockout mutation in the protease. We then expressed the revertant (double mutant) proteins under radiolabelling conditions and analyzed the assemblies produced using our standard biochemical assays (Figure 7). When the mutant proteins were expressed along with the HK97 protease, Prohead II was made by each of the mutants, as indicated by the prohead bands seen in the agarose gel shown in (a) lanes 3, 5 and 7, but each of the mutants also showed a prominent band of capsomers which was not observed in the wild-type control. The relative amount of Prohead II and capsomers produced was different for each double mutant, with the efficiency of prohead production (in order of amount of prohead produced) of WT > K178N/D231L > K178I/D231L > K178V/D231L. To our surprise, we could not isolate a Prohead I-like particle from any of the double-mutants when expressed in the absence of the HK97 protease (Figure 7 (b) lanes 4, 6, 8). Note that the wild-type gp5 with a protease knockout control produced a high yield of Prohead I and few, if any, capsomers (Figure 7 (b) lane 2).

Figure 7. Properties of pseudorevertants of D231L.

Analyses showed that D231L pseudorevertants assemble less efficiently than wild type and that Prohead I cannot be isolated from the revertants. Radiolabelled extracts from cells expressing wild-type or mutant proteins were separated into soluble or pellet fractions and analyzed for assembled protein using agarose gels in (a) and (b). Expression of pseudo-revertant (double mutant) mcp's (a) in the presence of the maturation protease, or (b) in the absence of the protease. (c) SDS gels of samples taken to compare the time course of cleavage of the pseudo-revertant (double mutant) mcp's to wild-type mcp cleavage under radiolabelling conditions during assembly in vivo using expression from plasmids.

These results suggest that the efficiency of prohead assembly in the double mutants is somewhat less than for the wild-type protein and further that the substitution of non-ionic for ionic interactions between residues 231 and 178 leads to a Prohead I that is so unstable that it is not detectable in the laboratory, but sufficiently stable in vivo to produce an effective yield of phages during an infection. Regardless of the instability of Prohead I in these mutants and the possibility that they may assemble more slowly, in the presence of an active protease, each mutant made a Prohead II which is stable and was easily detected in our plasmid/gel-based analyses. Our ability to detect the assembly of these unstable pseudoreverant Prohead I's via their conversion to a stable Prohead II in the presence of an active protease is an important observation. This supports our interpretation (above) that the D231 and K178 mutants which appear to make only capsomers in the presence of an active protease are simply unable to assemble into proheads, since any proheads made would be trapped as Prohead II by proteolysis, and argues against the alternative possibility that they do assemble into proheads, but that the proheads are too unstable to detect.

Discussion

The transient nature of the G-loop-E-loop contacts

While most, if not all, tailed bacteriophage capsids appear to be made from proteins which share a common protein fold^{41; 42}, no common underlying mechanism for producing capsids of the correct size and shape has been discovered. However, it is clear that any such mechanism must function at the early stages of assembly, which, in HK97, is during the assembly of Prohead I. We suggest that the forces specifying size and shape operate at these early stages and that much of the information controlling capsid shape is often erased during the maturation steps that occur subsequently - including proteolysis, scaffolding protein removal, and the conformational changes that occur during the expansion transformation, or during the analogous changes that take place without expansion in ϕ29 and the Herpesviridae ^{43; 44}. This means that a structure of a mature phage capsid, no matter how high the resolution, will not easily reveal how its shape was regulated. This was the case for HK97, where the structure of the mature head could not reveal the transient roles of residues 178 and 231 in assembly.

In a simplistic view of HK97 assembly, the charged residues D231 and K178 would be required to interact with their oppositely charged partners in order to orient the capsomer interfaces and facilitate correct assembly of preformed capsomers of exactly the correct size and shape (Figure 8 (a,b)). However, this view ignores the fact that it is not the charge of these residues that matters (see below). The same view also ignores the many other contacts, charged and otherwise, that comprise the capsomer-capsomer interfaces of HK97 proheads. These include a set of “anchoring” ionic interactions (or “staples” - those between E363 and R194 and between E344 and both R347 and R194)^{9; 45} that are found in all X-ray structures of HK97 from Prohead II⁹ to Head II¹ and therefore appear to be maintained from assembly to maturation, despite the large-scale conformational changes that take place (see Figure 8(c) and Supplemental movie S1).

Figure 8. D231-K178 interactions in HK97 proheads and assembly models.

(a,b) The assembly of HK97 proheads could occur via the assembly of pentons with hexons that are already in the correct conformation, but simply need to be oriented by the interactions of charged residues D231 and K178. This scheme, however, ignores the contributions of other charged residues which have also been shown to have important roles in assembly, including the 4 additional charged residues shown in panel (c). An alternative model for assembly is shown in panels (d) through (f), where a hexon in what is assumed to be a symmetric configuration adopts its characteristic asymmetric shape via changes induced by assembly. In this scheme, the shape changes of the subunits of the hexon are partly limited by contacts between residues D231 and K178. The same model in depicted in Supplemental movie S2.

Is it the sizes of residues 231 and 178 that count?

We have shown that the interactions between residues 231 and 178 are essential and can only be replaced by specific combinations that we have so far only been able discover by using genetic selection. The combination of residues (other than the wild-type pair, D-K) that are functional for assembly, L-N, L-I and L-V, all replace the original ionic contacts with contacts between pairs of residues that model building shows can comfortably bridge the space between the capsomer structural elements via van der Waals contacts (not shown). The conclusion that it is the steric and not the ionic nature of the interaction that is important is supported by sequence alignments of HK97-related capsid proteins in which the residues at the corresponding positions are more likely to be hydrophobic or neutral (a frequently occurring pair is L-A), rather than charged, and only rarely do oppositely charged pairs appear (see the alignment of HK97-related capsid protein sequences displayed at http://gigapan.com/gigapans/b806a9da80b4a66bcb1dd1b273d5b7a4/). The charged pairs that do occur in these positions are all D-K and include only a few very close relatives of HK97 and the capsid protein of phage XP10 and its few close relatives. In light of these observations, we think that for phage capsid proteins that utilize an HK97-like G-loop-E-loop interaction, it is not the charge, but the size that counts. This is emphasized by our results, above, in which we show that maintaining the charge, but increasing the size of these residues, in most cases leads to a complete failure of assembly beyond capsomers. The changing of the charge of these two residues (above) had effects which might be called gain-of-function, because they lead to drastically altered assembly, mostly into tubes. This altered assembly could easily be due to interactions of the charge-alterred residues with new partners found in their new enviroment, or due to unfavorable orientation of the interacting pairs, since salt bridges are known to have favored geometries.⁴⁶

What is the G-loop for and why is it conserved?

The widespread evolutionary conservation of the G-loop suggests that it plays an important role in capsid biology. Our results suggest that this conservation is driven by the need for the G-loops to participate in capsomer-capsomer interactions in a way that is essential for correct assembly. We speculate that the G-loop's major role is to limit one aspect of the capsid protein's search for the correct conformation during assembly, the local dihedral angle between a subunit and its neighbor on the adjacent capsomer. The rigidity of the connection of the G-loop to the backbone helix of the HK97 subunits can be seen when it is presented in space-filled representation (Figure 1 (b)). In this view it, can be seen that the G-loop appears tightly packed and firmly supported by the four preceding and three following helical residues. Thus, the G-loop appears to be a rigid extension of the backbone α-helix that is likely to be stiff enough that moderate twisting or rotation of parts of the subunit may be stopped or limited when a G-loop contacts a neighboring subunit. We suggest that the G-loop may serve exactly that kind of function during assembly - to allow a subunit to twist or flex while partly held in place by the anchoring salt bridges of the P-domain and elsewhere, until the G-loop limits the motion, and the subunit can eventually be locked into the correct conformation. Individual subunits of a hexamer or pentamer are certainly not independent, so changes in one subunit within a hexamer will influence its neighbors. This interdependence among capsomer subunits means that the ensemble could gradually change conformation in a programmed manner in response to binding to first one neighbor, then another, and so on until assembly is locally complete.

If capsomers behaved as rigid bodies, then the local dihedral angle specified by G-loop-to-E-loop contacts could be directly involved in size determination, with larger angles for larger diameter capsids and smaller angles for smaller capsids. If that were simply true, and the major factor in size determination, then the mutants we described above (D231E , K178R and D231E/K178R) would be predicted to possibly make smaller diameter assembled structures. Although these three mutants produced mostly capsomers (Figure 6(a)), mutant D213E did produce some smaller diameter capsid fragments (Figure 5 (a,b)). However, these structures failed to produce closed shells, suggesting that the relationship between these contact residues and capsid size is a complex one. Small and incomplete capsid fragments were also made by mutant D231A and other mutants produced long or bent tubes of various widths, which varied from mutant to mutant (Figure 5), further demonstrating that residues 178 and 231 play a direct role in specifying the form into which the HK97 capsid protein assembles.

There is ample reason to believe that capsomers do not behave as rigid bodies, given that the HK97 major capsid protein subunit has been shown to adopt a variety of related, but different conformations.^{2; 8; 9; 47; 48; 49} The HK97 subunit flexibility is portrayed in the movies included as Supplemental information for a previous publication⁹, and this one (see Supplementary movies). We have argued that the hexameric capsomers of HK97 may exist both in assembly-competent and assembly-incompetent forms^{39; 50}, but we have so far been unable to biochemically separate such forms. This may be because unassembled capsomers of HK97 major capsid protein are rather flexible in conformation, and their conformation may change as they interact with their neighbors in specific ways during assembly.

Conformational control of assembly

How could the type of interactions described above lead to the assembly of capsids of the correct size and shape? Applicable models proposed for regulating capsid geometry during assembly fall under two general headings: 1) local rule models^{51; 52}, where each subunit obeys abstract rules about how it can or cannot interact with its immediate neighbors in order to assemble correctly or 2) hexamer-type or hexamer-shape models^{53; 54}, where a limited set of hexamer shapes can only bind to each other in specific ways to produce a capsid of a specific size. Hexamer-shape models are attractive because the smaller T numbers only require 1, 2, or 3 hexamer shapes for T numbers up to 19, but hexamer-shape models require an unspecified mechanism by which hexamer shape can be regulated. We merge the two types of models by proposing that the bio-mechanical properties of interacting proteins within a hexamer have been genetically selected to change conformation in a programmed manner in response to binding to their neighbors. In this way, hexamers can act as both the hardware and the software for implementing a set of “local rules” to dynamically regulate the hexamer shapes and thus the shape and size of the assembling capsid. An example showing the kind of conformational change that may occur during assembly is shown in Figure 8 (df) and in Supplementary Movie S2. We show how an imaginary symmetric free hexamer of HK97 subunits might interact with other hexamers and pentamers during assembly. We envision that assembly is partly guided by the complementary interactions of the charged residues on the periphery of capsomers that are highlighted in the figure, and that the G-loop provides a sort of door stop that limits the extent of the subunit's conformational change when it contacts the adjacent E-loop.

What we are proposing is that propagating interactions similar to those which occur in many assembly pathways,^{55; 56; 57} where the order of assembly of multiple different proteins is controlled, are used in procapsids to regulate conformation of the same protein and its interactions to determine capsid size or T number. If such ordered interactions can produce something as complicated as the phage contractile tail^{58; 59} or the bacterial flagellum⁶⁰, then the related scheme that we suggest should be able to control the shape of a capsid. Berget⁶¹ and King⁶² have discussed such mechanisms in detail for sequential assembly of different proteins, which King called self-regulated assembly, “in which polymerization is regulated through the interactions of free subunits with an organized structure.”

The above model is purposely left vague in detail, to be filled out later, but as a working model it helps explain some of our current observations, and we think it provides a good framework for the continuing work with HK97 capsids. Under this model, tubes form when a structure initiates assembly and leaves at its growing end a self-propagating conformation of hexons that induce all subsequent capsomers that join to change into the tube-forming state, ad infinitum. As such, tube formation behaves as a prion-like disease of capsid protein that would compete with normal assembly and therefore must be strongly selected against in capsid protein evolution. It is usually assumed that HK97 assembles from preformed hexamers and pentamers, but the above model would work equally well during assembly from free subunits.

Homologs and potential analogs to the HK97 G-loop in other phages

Phage capsid proteins which possess a detectable G-loop homolog (See Figure 1 and Introduction) are likely to use it in the same manner as it is used in HK97, which, as proposed above, may be to control or guide assembly at the capsomer level. However, there are several additional protein models from X-ray and cryo-EM studies that clearly have variants of the HK97 fold that are distinctly different, each with its own set of additions and/or subtractions of structural elements. These include the Type 1 encapsulin protein (PDB ID: 3DKT)²⁵ from Thermatoga maritima, the Pyrococcus furiosus Type 2 encapsulin protein (PDB ID: 2E0Z), ⁶³ the mature capsid protein of phage ε15 from cryo-EM (PDB ID: 3J40),²⁰ the latest P22 procapsid model (PDB ID: 2XYY)¹², the T4 capsid vertex protein gp24 (PDB ID: 1YUE),¹⁴ the phage ϕ29 major capsid protein,¹⁶ the putative E. coli CFT073 prophage capsid protein subunit (PDB ID: 3BQW),¹⁵ and the putative Bordetella bronchiseptica prophage capsid protein (PDB ID: 3BJQ). With only one exception (2E0Z), each of these has an identifiable G-loop or additional structural element which is arranged in a manner that suggests it could play a role in assembly similar to the role of the G-loop in HK97 by contacting subunits in adjacent capsomers to assist in regulating assembly.

The D-loops of P22 appear to behave similarly to the G-loops of HK97. The P22 procapsid model (PDB ID: 2XYY)¹² has protrusions from P22's “insertion domain” (I-domain, residues 260-279, also called the ED or extra domain) called D-loops which are similarly located to a G-loop in three dimensions and appear to make contacts with the I-domains of subunits in adjacent capsomers. The contacts between the D-loops and the neighboring subunits are broken during P22 expansion in a similar manner to that seen for the HK97 G-loop-E-loop contacts. This is illustrated in Supplementary Movie S3, where expansion of part of the P22 capsid is modeled (using PDB ID: 2XYY and the P22 mature capsid model PDB ID: 2XYZ)¹¹ with D-loops colored in pink. Mutations that result in the formation T=4 instead of the usual T=7 P22 shells map to residue 285⁶⁴, in a helix that is directly connected to the D-loop in the current P22 models discussed above, suggesting that changes at residue 285 may change capsid size by altering D-loop geometry.

Conclusion

Structural studies of HK97 capsids revealed an unusual feature called the G-loop, which appears to be conserved in closely related proteins. We noted that a specific residue in the G-loop makes a transient contact with another residue in the E-loop of the adjacent capsomer. We made mutations that altered the unusual feature and the contacting residues, studied their consequences, which included abnormal assembly or lack of assembly, and selected for revertants that could restore function. The detailed results, combined with the available structural information, led us to look at the bigger picture and speculate about the role of G-loop interactions in regulating HK97 assembly, and then to look further and see if our speculations might apply to more distantly related capsid proteins. We find that other capsid proteins in the HK97-fold family have a variety of unexplained structural features that occur in positions that would enable them to regulate assembly in a manner related to what we propose for the HK97 G-loop.

Materials and Methods

Bacteria, bacteriophage and plasmids

E. coli strains Ymel (mel-1 supF58) and LE392 (F^-, metB1, trpR55, lacY1, galK2, galK22, supE44, supF58, hsdR514) were used for propagating HK97 wild type and amber mutants. All HK97 lysogens were derived from the strain BW25113 (Δ(araD-araB)567, ΔlacZ₄787(::rrnB-3),-λ-, rph-1, Δ(rhaD-rhaB)568, rrnB-3, hsdR514 ). HK97 revertants were selected by plating on Ymel. T7 promoter expression strain E. coli BL21(DE3)/pLysS⁶⁵ was used for plasmid complementation tests and for production of proheads from plasmids. E. coli DH10B was used for molecular cloning steps. Plasmid pVB, a protease-knockout plasmid (expressing HK97 gene 5), was used to produce Prohead I and plasmid pV0 (expressing HK97 gene 4 and gene 5) was used to produce Prohead II⁴. Plasmid pV0 was used for construction of variants at residue 231 and 178 in gene 5, D231L, D231A, K178E, K178L, and a deletion of the G-loop, Δloop, using site-directed mutagenesis PCR. Mutations were made by using primer pairs 5’CAGCCCGTTCAGCAGCTGGCCC and 5’CGGGCTGAACAAAGTGGCAACC for Δloop, 5’GGTACCGGTCTCAACCTGGA and 5’TCCAGGTTGAGACCGGTACC for D231L, 5’CCGGGGCTAACCTGGAAGG and 5’CCTTCCAGGTTAGCCCCGG for D231A, 5’GATATCACCTTCAGCGAACAAACCGCGAACGTG and 5’CACGTTCGCGGTTTGTTCGCTGAAGGTGATATC for K178E, 5’GATATCACCTTCAGCCTGCAAACCGCGAACGTG and 5’CACGTTCGCGGTTTGCAGGCTGAAGGTGATATC for K178L. The K178E and K178L mutations were made by B. Baros (Brandi Lynn Baros, Ph.D Thesis, 2002, University of Pittsburgh, Pittsburgh, PA). A translationally silent restriction site polymorphism was created at or near each target residue codon in order to simplify the screening of potential mutants, lysogens, and revertants for the presence of wild-type sequences. Candidate mutant plasmids were sequenced to ensure that only the desired changes were present.

Additional mutations of HK97 gp5 codons 231 and 178 were made using a variation of a PCR-based method that uses phosphorylated primers⁶⁶, using Phusion polymerase (New England Biolabs) instead of Pfu and also using first-round PCR products purified by gel electrophoresis and pressure-extrusion⁶⁷. Primer pairs for D231E were D231Er (5’P-gttctccccggtaccgtcgccgt) plus HK155f (5’gtgtttaccaataacgccgacgt), and HK233f (5’P-ctggaagggctgaacaaagt) plus rmp5 (5’cccctccatcatgagcc). For making D231K, primer D231Kr (5’P- gttcttcccggtaccgtcgccgt) was substituted for D231Er and other primers were the same. Primer pairs for K178R were K178Rr (5’P-gcggctgaaggtgatatccga) plus upstr2for ( 5’cggaaacgaagcacaaataaa), and HK179f (5’P-caaaccgcgaacgtgaagac) plus RP1 (5’ccttccaggttatccccgg). For making K178D primer K178Dr (5’P-atcgctgaaggtgatatccgat) was substituted for K178Rr and other primers were the same. After second-round PCR reactions⁶⁶, PCR products were purified, cut with restriction enzymes (SacI and DraIII for K178 mutants, and BtgI and DraIII for D321 mutants) and the mutant-containing restriction fragments were used to replace the corresponding wild-type segments of HK97 wild-type expression plasmid pV0-SacI. Candidate mutant plasmids were screened by DNA sequencing of the PCR-generated parts.

Construction of an HK97 lysogen recombineering strain

Recombineering techniques were used to create the strains needed to introduce lethal missense mutations into the HK97 lysogen by homologous recombination^{40; 68}. The host for all recombineering experiments was E. coli strain BW25113 containing the plasmid pKD46, which expresses the phage λ Red recombination proteins under arabinose control⁶⁸. The galK positive/negative selection system was used to facilitate selecting recombinants⁶⁹. An HK97 lysogen of strain BW25113 (pKD46) was made by infecting with HK97 at a high multiplicity and tested for homo-immunity by cross-streaking with the λ clear plaque variant λb2cI. Next, the galK gene was deleted via recombineering after inducing BW25113::HK97 (pKD46) with arabinose as previously described⁶⁸. The ΔgalK linear recombination substrate was made from chromosomal ΔgalK DNA from the galK deletion strain SW102 described by Warming⁶⁹ using primers in galT (gaccgacgcccagcgcagcgatctggcgctggcg) and in galK (gcaccgtcgcgccgagattacccgggaagccctg). Induced cells were transformed with ΔgalK DNA, allowed to outgrow and selected on minimal plates containing 0.2% 2-deoxy-galactose (DOG)^{69; 70}. The deletion was confirmed by colony PCR with the same primer pair and named DT002.

Next, a segment of HK97 gene 5 in DT002 was replaced with galK. The galK gene on plasmid pgalK⁶⁹ was amplified with primers that provide ~60 bases of homology to the HK97 chromosome at both ends (shown here in lower case); the primers used were HRgalKf-1 (gcacaaataaacgctctgcttcagagcatcaaatctttcccttctaacttaggCCTGTTGACAATTAATC ATCGGCA) and HRgalKr (gggttcaggacgataccggaagcgctgaactcagactcggtcacctGTCAGCACTGTCCTGCTCCT T). This primer pair produces a 1331 base pair fragment which was used to replace the HK97 DNA from codon 224 in gene 4 through codon 264 in gene 5 with galK. The amplified DNA was recombined into the ΔgalK HK97 lysogen (DT002) as described above and selected on galactose minimal salts plates⁶⁹. The deletion was confirmed by colony PCR using primers that bind to regions that flank the deletion in the HK97 chromosome, and the strain was named DT005. The DT005 HK97 lysogen contains the galK replacement of the first 264 codons of gene 5 in the lysogen and was used to create HK97 lysogens that have lethal mutations in gene 5 for use in selecting revertants and pseudorevertants as described below.

Recombineering substrate DNA containing mutant gene 5 (with the D231L substitution) was PCR-amplified from a plasmid with the mutation using primers LP1 (GGTGACGGGAAGCAGGGG) and gp6-r1 (GTCGTCCCTGTCGTCTTCCTC), gel purified, and used directly. DT005 was cultured in LB with 100 μg/ml ampicillin and 10 mM L-arabinose (to induce λ Red gene expression) at 30°C until cells reached A₅₅₀ ~ 0.4 as measured using a Spectronic 20 spectrophotometer (Bausch & Lomb). Cells were harvested, chilled on ice for 10 min, centrifuged at 6700 rpm for 10 min at 4°C, washed three times with ice-cold distilled water and resuspended in a small volume of chilled water. The purified linear DNA containing the HK97 gene 5 D231L mutation was introduced into the treated cells using electroporation. The cells were diluted into LB medium with 100 μg/ml ampicillin plus 10 mM L-arabinose and incubated at 30°C for ~3 hours allowing cells to continue to express λ Red genes and to segregate chromosomal copies of the selectable genes. Cells were washed three times with M9 or M63 minimal salts, resuspended in a small amount of minimal salts and spread on M63 minimal plates with DOG for selection for loss of the galK marker. Recombination efficiency was around 10^-6 or lower. Portions of HK97 gene 5 from candidate colonies were amplified by colony PCR and screened for the presence of the restriction polymorphism that was included in the original mutagenesis step to mark the mutated lysogen (a new BsmAI site for the D231L mutation).

Expression and radiolabeling of HK97 proteins

Wild-type and mutant HK97 proteins encoded on plasmids were expressed and labeled with ³⁵S-Met in the host BL21(DE3) pLysS⁶⁵ using protocols similar to those described previously⁴ in medium supplemented with 50 μg/ml ampicillin and 25 μg/ml chloramphenicol. The medium normally used was M9 with 1% Difco Methionine Assay Medium, but LB was also used for several experiments with comparable results. Cells, 50 μl of a turbid culture) carrying the target plasmid, were diluted into 5 ml medium and held at 30°C overnight without aeration. The next day the culture was moved to a 37°C water bath and aerated until its A₅₅₀ was ~0.4. IPTG was added to 0.4 mM to induce the target gene expression. After incubation for 30 min, 200 μg/ml rifampicin was added to inhibit host protein synthesis. After 15 min, 1 ml was moved into a tube and radiolabeled by the addition of 1 μl ³⁵S-Met (10 mCi/ml). At intervals after ³⁵S-Met was added (0 min, 5 min, and 45 min), 100 μl samples were precipitated with 10% TCA. The TCA-precipitates were washed with acetone, collected, dried, resuspended, denatured in SDS sample buffer and analyzed using SDS polyacrylamide gels. At 45 min after label addition, the remainders of the labeled cultures were chilled on ice for 10 min, pelleted by brief centrifugation, and suspended in 500 μl lysis buffer. Triton X100 was added to 0.1% and the samples were warmed to 24°C for 5 min to initiate bacterial lysis. Lysates were chilled, 7.5 mM MgSO₄ and 20 μg/ml DNaseI were added. After being held at ~26° C for 10 min for DNA digestion, the lysates were chilled and insoluble material was removed by centrifugation at 12 kRPM for 10 min at 4°C. The lysate supernatants were saved, and the pellets were suspended in 500 μl of TKG50 buffer. All samples were kept at 4°C until they were analyzed on native agarose and native polyacrylamide gels.

Small scale (mini-prohead) preparations used auto-induction methods⁷¹ to express HK97 capsid proteins in 1.5 ml cultures containing TYM5052 medium⁷¹ supplemented with 50 μg/ml ampicillin and 25 μg/ml chloramphenicol grown at 37°C with rapid shaking at 300 rpm for 24 hrs in 18 × 150 mm tubes. The host cells were BL21(DE3)⁶⁵ that also contained plasmid pLysS. Cultures were collected by centrifugation, suspended in 300 μl 50 mM Tris-HCl pH 8 with 0.008 M EDTA, and lysed by freezing and thawing after the addition of Triton-X100 to 0.15%. After slow thawing, MgSO₄ was added to 25 mM along with DNaseI to 15 μg/ml and the suspension was incubated at 28°C for ~5 min for DNA digestion. The resulting crude extract was separated into three fractions: pellet (containing insoluble, large or aggregated material) , supernatant or lysate (prohead, capsomers, etc.) and a PEG precipitated supernatant fraction (containing proheads, heads, etc., but depleted of capsomers and host proteins). Centrifugation at 14,000 rpm for 10 min produced pellet and supernatant fractions; the pellet was suspended in 20 mM Tris-HCl pH 7.5 with 40 mM NaCl. Half of the supernatant was precipitated by adding an equal volume of 12% (w/v) PEG 8000 with 1 M NaCl, leaving the mixture on ice for 20 min, and then collecting the pellet by centrifugation as in the previous step. The PEG supernatant was discarded and the PEG pellet suspended in 20 mM Tris-HCl pH 7.5 with 100 mM NaCl. All samples were stored on ice until analyzed.

Agarose gel analysis of HK97 head-related structures

Native agarose gel electrophoresis was used as a quick assay for large capsid protein assemblies. Samples were electrophoresed in submarine mini-gels using TAMg buffer (40 mM Tris base, 20 mM acetic acid pH 8.1, and 1 mM magnesium sulfate). Samples were prepared by mixing 9 parts sample with 1 part of dye/glycerol solution (50% (v/v) glycerol, 0.025% (w/v) bromphenol blue, 0.025% (w/v) xylene cyanol XFF). After electrophoresis, gels were fixed and stained for protein using Coomassie Brilliant Blue R250 and dried, if necessary, after staining, for autoradiography of radiolabeled proteins.

Electron microscopy

Samples were applied to glow-discharged 400 mesh carbon/formvar grids (Ted Pella, Redding, CA), rinsed with water, stained with 1% uranyl acetate and blotted dry. Samples were visualized using a Morgagni 268(D) EM (FEI, Eindhoven, Netherlands) at 56,000X magnification with images recorded digitally using a side-mounted ORCA camera (Hamamatsu, Bridgewater, NJ). Mild unsharp mask filtering was used to enhance contrast.

SDS polyacrylamide gel analysis

SDS-polyacrylamide gel electrophoresis conditions were modified from Laemmli⁷². A low-cross-linker acrylamide stock containing 33.5% (w/v) acrylamide and 0.3% (w/v) methylene bis acrylamide⁷³ was used to make most gels as an alternative to the standard⁷² 30%:0.8% stock. All samples were heated in boiling water for 2.5 min after mixing with ¼ volume of SDS sample buffer—the 4x sample buffer normally used contained 0.25 M Tris-HCl, pH 6.8, containing 8% (w/v) SDS, 20% 2-mercaptoethanol, 40% glycerol, and ~0.025% (w/v) bromphenol blue.

Molecular Graphics

SPDBV was used to create and visualize HK97 capsid protein oligomers. SPDBV combined with POV ray tracer (http://wiki.povray.org/) and variant MegaPOV (http://megapov.inetart.net/) were used for rendering figures and images for movies. Chimera⁷⁴ was used for viewing density maps and Cα traces. For making movies, all PDB chains were modified to have the same length for ease of manipulation and for successful interpolation using The Morph Server⁷⁵ (http://molmovdb.org). For morphing from Prohead II (PDB ID: 3E8K) to Head II (PDB ID: 1OHG), model chains were all modified to all start at residue 128 and the truncated E-loop used for the Prohead II structure⁹ was used to replace the full-length E-loop in Head II models. The symmetric hexon model was made from 6 copies of the F chain from PDB ID: 3E8K (Prohead II). P22 morphing was done using the morph.tcl add-on for VMD⁷⁶, using procapsid model PDB ID: 2XYY and a version of the mature capsid model PDB ID: 2XYZ truncated to have the same number of residues as 2XYY.

• Ten residue-long gycine-rich G-loops protrude from the HK97 capsid surface

• Transient G-loop contacts with adjacent capsomers are essential for assembly

• G-loop - E-loop contacts control the shape of HK97 capsid protein assemblies

• The size, not the charge of G-loop - E-loop contact residues is important

• G-loops and other G-loop-like protrusions are common features in capsid proteins