How phage L capsid decoration protein distinguishes between nearly identical binding sites on an icosahedral virion

The major coat proteins of most dsDNA tailed phages and viruses form capsids by a mechanism that includes active packaging of the dsDNA genome into a precursor procapsid, and subsequent expansion/stabilization of the capsid. Packaging of dsDNA genomes accrues ∼10 - 60 atm of internal pressure. Phages and viruses have evolved diverse strategies to strengthen their capsids, such as non-covalent binding of auxiliary “decoration” proteins. The P22-like phages are paradigms for understanding the biophysics of capsid assembly and maturation. For example, phage L, Dec protein that has a highly unusual binding strategy where precisely distinguishes between nearly identical regions of the capsid. We combine cryo-electron microscopy and three-dimensional image reconstruction at near atomic resolution, with molecular dynamics techniques to discern the structure of native phage L particles. We used NMR to determine the structure/dynamics of Dec in solution. Key regions of coat and Dec that modulate the selectivity of Dec binding are elucidated.


50
Viral icosahedral capsids are formed from multiple copies of a single or a few types of 51 coat proteins that encapsidate the genome. Because of the symmetry of capsids that have 52 more than 60 coat subunits, genetically identical proteins must assemble into capsid sites 53 with different quasi-equivalent conformations, such that some coat proteins are in hexameric 54 conformations and some are in pentameric conformations [1]. The resulting capsid shells 55 often have additional decoration proteins bound to their exteriors that stabilize the virion 56 against environmental insults. How decoration proteins bind to specific sites on quasi-57 equivalent capsids is poorly understood.

117
We report the capsid structure of the phage L virion, and elucidate the mechanism of probed by site-directed mutagenesis. Our findings reveal that Dec has a novel protein fold 6 As a first step towards understanding the mechanism by which Dec binds to phage L, 136 we determined a near-atomic level resolution structure of native phage L capsids using cryo-137 EM ( Figure 1A). The resolution of the entire density map was determined to be 4.2 Å (see 138 Supplementary Materials, Figure S1). Both the overall protein structure, as well as the 139 specific contact points for attachment of Dec trimers to the coat protein subunits are clearly 140 discernable near quasi-three-fold symmetry axes between hexons. As anticipated from 141 previous studies [17,18], Dec very sparsely occupied true three-fold symmetry sites in phage 142 L virions, as there was only a hint of very weak density observed at these positions, and no 143 Dec occupancy was observed at quasi-three-fold sites between hexons and pentons.

151
As an initial guide for accurately fitting the phage L coat protein into the cryo-EM 152 density, we used the most recently deposited structure of a P22 coat protein asymmetric 153 unit (PDB ID: 5UU5; [26]), since phage L and phage P22 coat proteins are nearly identical, 154 with only four amino acid differences [17]. Upon initial docking, we found deviations between 155 the phage L capsid density and the P22 coat protein asymmetric unit. Therefore, to optimally 156 fit the cryo-EM density we built a model of the phage L coat protein lattice where each 7 accounted for during modeling; the R101H side chain is located in the spine helix pointing 162 towards the capsid interior, I154L is located in the A-domain towards the hexamer center,

163
M267L is in the I-domain, adjacent to but not interacting with Dec, and A276T in also located 164 in the I-domain but on the distal end pointing towards center of the hexamer ( Figure S2D).

165
Overall, there were minor differences between the optimally fit P22 and phage L capsid

181
The Dec protein fold was determined by first using homology modeling (see 182 Materials and Methods for details) followed by a fully atomistic refinement fit to the 183 segmented cryo-EM density using MD-based fitting ( Figures 1C, and S4). In brief, the N-184 terminal domain was modeled using an OB fold (described in more detail below) and the C-185 terminal domain with a three-stranded b-helix domain. The presence of an OB fold was 186 identified by homology and a template-based model was initially docked into the cryo-EM 8 density envelope. Studies using NMR also revealed an OB fold from Dec monomers, not 188 bound to the capsid (see NMR section below). The final trimeric Dec structure was refined 189 using MD-based flexible fitting within the cryo-EM density envelope, using the NMR 190 constraints to generate the final Dec structure.

191
The N-terminal domain of Dec (residues 10-88) was well defined by the cryo-EM 192 density, but residues 89-134 were predicted with a lower confidence owing to the flexibility 193 and lower resolution within that portion of the structure (Figures S3, S4). As we previously

201
To further investigate the properties of Dec, we characterized the unassembled 202 protein in the absence of capsids by solution NMR. Initial NMR spectra of Dec at neutral pH 203 gave extremely broad lines, with poorer signals than expected for a 43 kDa trimer, which 204 was likely due to aggregation. We were able to achieve reproducibly good samples that 205 gave NMR spectra with sharper lines using a protocol [30] in which Dec is acid-unfolded by 206 lowering the pH to 2, followed by a refolding step induced by adjusting the pH to 4.

207
Subsequent characterization of the protein using size exclusion chromatography, native-gel 208 electrophoresis, and 15 N NMR relaxation measurements ( Figure S5A) showed that the 209 unfolding/refolding protocol converts purified Dec from a trimer to partially folded monomers 9 The good quality of the NMR data obtained following the acid-unfolding/refolding 214 protocol, enabled us to obtain nearly complete (98%) NMR assignments for the monomeric

237
In general, the cores of OB-fold b-barrels are consolidated by three layers of hydrophobic 238 10 residues [24,34]. This arrangement is also present in the Dec structure (not shown). The 239 canonical role of the helix aOB is to provide a 'hydrophobic plug' for the bottom hydrophobic 240 layer of the b-barrel [24,34]. Residue V55 appears to serve this role in Dec. In OB-fold 241 proteins the orientation of the helix aOB is more variable than that of the b-barrel [35,36], 242 and in Dec, the aOB helix extends almost directly between strand b3 and b5, with the helix 243 axis in the plane of the b1-b3 meander, rather than below this structure ( Figure 3B).

254
To further characterize the dynamics of Dec, we collected NMR 15 N-relaxation data 255 ( Figure S5) and analyzed the dynamics of the monomeric protein in terms of the "Model-

256
Free" formalism. The local backbone mobility of the Dec monomers is summarized in Figure   257 4. The N-terminus from residues 1-11 and the C-terminus from residues 90-134 have small

278
To assess the role that specific coat protein surface residues play in modulating Dec 279 binding, the following five coat protein residues in potential close contact with Dec were

12
We noted that coat proteins occupying different local conformations (as a result of 290 the quasi-equivalent capsid lattice) contribute different residues to the binding interface 291 ( Figure 5A, Movie S3). For example, residues E81, P82, and R299 from the coat protein 292 subunits that form the quasi-three-fold symmetry axes (grey subunits in Figure 5) contact 293 Dec, whereas residues P322 and E323 contact Dec from adjacent and overlapping coat 294 protein subunits (black subunits in Figure 5). While their closeness to Dec clearly indicates 295 that it interacts near these coat residues, there are no obvious salt bridges between the coat

304
Therefore any changes in Dec occupancy in the variants are likely due to a disruption of the 305 binding interface, rather than protein folding aritfacts.

306
Among the five coat protein changes, coat amino acid substituions E81R and E323R 307 attenuated but did not completely abolish Dec binding ability (Figure 5 B,C). Therefore, we

316
The close contact region between Dec and coat can be thought of as including two 317 binding sites. One includes coat residues E81 and P82 along with their closest interacting 318 Dec residues K30, Y31 and Y49 ( Figure 5D, "site 1"). For example, Dec residue Y49 is 319 pointed directly towards the side chain of coat residue E81. The other patch includes Dec 320 residues Y71 and E73 and coat residues P322, E323 and R299 ( Figure 5D, "site 2"). Overall,  3D capsid structure. Coat protein residue P82 is at a lower virion radius than E81 and likely 333 too far from the Dec residue Y49 to affect binding when it is altered. Coat protein change 334 R299E did not alter binding, and this is consistent with the fact that when the closest 335 adjacent Dec residue Y31 is changed to an alanine there was also no effect on binding.

336
Lastly, coat residue P322 forms an interface with two Dec residues, Y71 and E73, and 337 single subsitutions at any of these positions have no effect on binding. Even though the 338 cryo-EM density is strong in this region indicating protein:protein contact, electrostatic 339 interactions may not be the driving force at site 2, but instead weaker interactions such as 340 14 van der Waals could contribute to binding. Further experiments will be required to fully 341 characterize any subtle changes in binding affinity of the variants that did not obviously 342 decrease binding saturation here.

343
Given that the binding interface is comprised of coat protein subunits in two distinct             contrast-transfer function, as described [58].   with the same protocol as before, allowing the C-terminal part of the trimer to move 540 separately from the N-terminus to find the best fit to the EM density. We note, that while the 541 model for the N-terminal part of Dec is supported by NMR data and a good fit to the 542 relatively high resolution EM density, the model for the C-terminus is speculative due to a 543 lack of high-resolution experimental data.

22
The initial Dec model was combined with six capsid proteins at the quasi-three fold 545 symmetry axis, and re-optimized via energy minimization to improve the fit at the interface 546 between Dec and the capsid proteins. The optimized Dec model was further then combined 547 with a larger capsid complex covering 1/8 th of the capsid. A copy of Dec was placed at each 548 quasi-three fold symmetry axis and an additional copy was placed at the true three-fold axis 549 as there was also faint EM density at the latter sites. The combined capsid-Dec complex

905
Chains of each trimer are rainbow colored from N-terminus (blue) to C-terminus (red).

913
We generated models for nine complete asymmetric units in order to completely fill the cryo-

914
EM density of one eighth of the total capsid ( Figure S2 A-C). Figure S2 indicates where 915 each of these asymmetric units are located and corresponds to the data in Table S1. We 916 37 carried out the same analysis with respect to the P22 structure based on the PDB structure 917 (5UU5).   Table S1). We carried out the same analysis with