Cryo-EM structure of the bacteriophage T4 isometrichead at 3.3-Å resolution and its relevance to theassembly of icosahedral virusesZhenguo Chena,1,2, Lei Suna,1,2, Zhihong Zhangb,1,3, Andrei Fokinea, Victor Padilla-Sanchezb, Dorit Haneinc, Wen Jianga,Michael G. Rossmanna,4, and Venigalla B. Raob,4

aDepartment of Biological Sciences, Purdue University, West Lafayette, IN 47907; bDepartment of Biology, The Catholic University of America, Washington,DC 20064; and cBioinformatics and Structural Biology Program, Sanford-Burnham-Prebys Medical Discovery Institute, La Jolla, CA 92037

Edited by Ming Luo, Georgia State University, Atlanta, GA, and accepted by Editorial Board Member Peter Palese August 18, 2017 (received for review May22, 2017)

The 3.3-Å cryo-EM structure of the 860-Å-diameter isometric mu-tant bacteriophage T4 capsid has been determined. WT T4 has aprolate capsid characterized by triangulation numbers (T numbers)Tend = 13 for end caps and Tmid = 20 for midsection. A mutation inthe major capsid protein, gp23, produced T=13 icosahedral cap-sids. The capsid is stabilized by 660 copies of the outer capsidprotein, Soc, which clamp adjacent gp23 hexamers. The occupan-cies of Soc molecules are proportional to the size of the anglebetween the planes of adjacent hexameric capsomers. The anglebetween adjacent hexameric capsomers is greatest around thefivefold vertices, where there is the largest deviation from a pla-nar hexagonal array. Thus, the Soc molecules reinforce the struc-ture where there is the greatest strain in the gp23 hexagonallattice. Mutations that change the angles between adjacent cap-somers affect the positions of the pentameric vertices, resulting indifferent triangulation numbers in bacteriophage T4. The analysisof the T4 mutant head assembly gives guidance to how othericosahedral viruses reproducibly assemble into capsids with a pre-determined T number, although the influence of scaffolding pro-teins is also important.

bacteriophage T4 | isometric head | virus capsid assembly | triangulationnumbers | capsid decoration proteins

Tailed bacteriophages are widely distributed in nature andprobably are the most abundant organisms on the planet (1,

2). Phage capsids are extremely stable and able to sustain highinternal pressure exerted by their tightly packed genomes (3, 4).Bacteriophage T4 belongs to the Myoviridae family of phages (5,6). Its 172-kbp DNA genome is packaged to near-crystallinedensity, resulting in an internal capsid pressure of ∼20 atm(7–9). T4 serves as an excellent model to elucidate the mecha-nisms of virus assembly and DNA packaging (10). T4 also pro-vides a valuable biotechnology platform for diverse applicationssuch as vaccine design and gene therapy (11–14).The T4 capsid, or head, is a 1,200-Å-long and 860-Å-wide

prolate icosahedron (15, 16). The capsid has a unique “portal”vertex to which a ∼1,400-Å-long contractile tail is attached (17,18). The tail is terminated by a baseplate (19–21) to which arebound six long tail fibers (18). During infection, the tail binds toan Escherichia coli cell and creates a conduit for injection of thephage DNA into the bacterial cytoplasm (22). The internalpressure then drives the DNA out of the capsid, into the cell,through the tail (3).The T4 head, tail, and fibers assemble by separate pathways

(5). They then join together to form the infectious virion. Thecapsid assembly proceeds via formation of a precursor, called aprohead, which contains an inner scaffolding core and an outershell (15). The prohead formation is initiated by the dodecamericprotein gene product (gp) 20, which makes the unique portalvertex of the prohead (23). Gp20 nucleates the assembly of theinner scaffolding core made by the major core protein, gp22, the

prohead protease gp21 (24), and other proteins. The major capsidprotein gp23 coassembles with the core to form the outer shell ofthe prohead. Pentamers of the vertex protein, gp24, form theother 11 vertices of the prohead (25). The gp23 and gp24 proteinshave approximately 22% sequence identity and likely have evolvedfrom a primordial gene via gene duplication.When the prohead assembly has been completed, the gp21

protease cleaves the inner core proteins (5, 15). Most of thecleavage products escape from the prohead to create space forDNA. In addition, the protease cleaves off 65 N-terminal resi-dues from the major capsid protein, gp23, and 10 N-terminalresidues from the vertex protein, gp24, producing gp23*(48.7 kDa) and gp24*, respectively. The genomic DNA is thenpackaged into the empty prohead through the portal vertex by anATP-driven packaging motor containing five subunits of themotor protein, gp17 (9). The gp23*–gp24* shell of the proheadundergoes large conformational changes, resulting in the ex-pansion of the capsid volume by ∼50%. In vivo, the expansion

Significance

The WT bacteriophage T4 head is prolate with icosahedralT=13 ends closing the central cylindrical section. The structureof a T=13 icosahedral head assembly, in which the major capsidprotein of T4 phage had a single mutation at a residue in theinterface between neighboring subunits, has been determinedto 3.3-Å resolution. The structure shows the extensive networkof interactions between the major capsid proteins gp23 andthe vertex proteins gp24 and the stabilizing Soc proteins. Weshow that the angle between hexameric and pentameric cap-somers helps to determine the triangulation number (T num-ber) and suggests how other viruses assemble consistentlywith a predetermined T number.

Author contributions: M.G.R. and V.B.R. designed research; Z.C., Z.Z., and V.P.-S. per-formed research; W.J. contributed new reagents/analytic tools; D.H. enabled some ofthe data collection; L.S. and A.F. analyzed data; Z.C., L.S., A.F., M.G.R., and V.B.R. wrotethe paper; and M.G.R. and V.B.R. supervised research.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. M.L. is a guest editor invited by the EditorialBoard.

Data deposition: The cryo-EM map of the isometric T4 capsid has been deposited in theElectron Microscopy Data Bank (ID code EMD-8661). The atomic coordinates have beendeposited in the Protein Data Bank, www.pdb.org (PDB ID code 5VF3).1Z.C., L.S. and Z.Z. contributed equally to this work.2Present addresses: The Fifth People’s Hospital of Shanghai, Fudan University, Shanghai201100, People’s Republic of China; and Institute of Biomedical Sciences, Fudan Univer-sity, Shanghai 200032, People’s Republic of China.

3Present address: Huaihai Institute of Technology, Lianyungang 222005, People’s Republicof China.

4To whom correspondence may be addressed. Email: mr@purdue.edu or rao@cua.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1708483114/-/DCSupplemental.

E8184–E8193 | PNAS | Published online September 11, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1708483114

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

happens during DNA packaging. However, these processes canbe decoupled (26). The outer surface of the expanded capsid isdecorated by the small outer capsid (Soc) protein and the highlyimmunogenic outer capsid (Hoc) protein (27). Soc helps tostabilize the capsid against extremes of pH and temperature,whereas Hoc probably assists in finding and infecting the bac-terial host (28, 29). After the completion of genome packaging,the portal vertex is sealed by the assembly of gp13–gp14 complex(i.e., “neck”), which also creates the attachment site for the in-dependently assembled phage tail (30).The structure of the mature T4 head was previously studied by

using cryo-EM (16, 23, 31, 32) (Fig. 1 A and C). The best reso-lution that had been achieved for the prolate head was approx-imately 10 Å (23). These earlier studies showed that the majorcapsid protein gp23* is organized into a hexagonal lattice char-acterized by triangulation numbers Tcap = 13 laevo for the ico-sahedral caps and Tmid = 20 for the elongated midsection (16).The Hoc monomers bind at the centers of the gp23* hexamericcapsomers, whereas Soc molecules bind at the interface betweenadjacent gp23* hexamers and clamp neighboring capsomers to-gether (29). The T4 head contains 155 hexamers of gp23*,12 copies of gp20 (61 kDa) at the special portal vertex, 11 pen-tamers of gp24* (44 kDa) at the remaining fivefold vertices,155 copies of Hoc (40.4 kDa), and 870 copies of Soc (9.1 kDa).The structure of the T4 vertex protein gp24 was determined by

using X-ray crystallography (25). The structure of gp24 has a fold

similar to the capsid protein of phage HK97 (33). The HK97 foldis a building block for the capsids of tailed dsDNA phages (34) aswell as for herpesviruses (35). The crystal structures of Soc (29)and Hoc (28) from T4-like phages have also been determined,although the Hoc structure did not include the C-terminal do-main, which binds to the capsid surface. In addition, an atomic-resolution cryo-EM structure of the dodecameric T4 portalprotein, gp20, has been determined (23). However, the structureof the major capsid protein gp23 had not been determined be-fore the present study because gp23 tends to aggregate whenproduced recombinantly, making it difficult to crystalize.Mutations in the T4 prohead proteins can change the length of

the capsid, producing mixtures of prolate WT capsids, icosahe-dral “isometric” or “petite” capsids, intermediate-length heads,and/or “giant” heads of various lengths (15, 36–38). In this study,we used a mutant T4 phage that produces predominantly emptyisometric heads. These isometric particles have icosahedralsymmetry except for the unique portal vertex occupied by thedodecameric portal protein gp20. The gp23* protein in theseisometric heads forms a hexagonal lattice characterized by thetriangulation number T = 13 laevo. Here we describe the cryo-EM structure of the isometric T4 head at 3.3-Å resolution. Theelectron density map established the atomic-resolution structuresof the gp23*, gp24*, and Soc molecules in the capsid environ-ment. Analysis of interactions between these subunits shows howa highly stable virus capsid can be assembled and how it can

Fig. 1. Structure of the T4 prolate (A and C) and isometric (B and D) heads. Gp23*, gp24*, Hoc, and Soc are colored blue, magenta, yellow, and white,respectively. Note that Soc is absent around the gp24* pentamers. In C and D, the gp23* protein has been removed for clarity.

Chen et al. PNAS | Published online September 11, 2017 | E8185

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

withstand internal pressure caused by the tightly packaged viralgenome. The structure also suggests the factors that determinethe Tend and Tmid values that describe the quasi-symmetry ofvirus capsids.

Results and DiscussionStructure Determination. Of several T4 phage mutants, the17amNG178.23pt21–34c mutant produced the largest fraction(approximately 80%) of isometric capsids (Fig. 2A). This T4 mutantcontained two mutations. The first of these mutations was A275T ingp23 that resulted in the production of isometric capsids togetherwith approximately 20% of prolate and intermediate-length heads.The second of these mutations was an amber mutation in thepackaging motor protein gp17 that resulted in the production ofDNA-free “empty” capsids when the phage was grown on a non-suppressor E. coli strain. The isometric capsids produced by thisdouble-mutant phage were assembled via the initial formation ofisometric proheads, followed by proteolytic cleavage of the pro-head’s inner scaffolding core, of gp23, and of gp24. The cleavedproheads remained empty because they could not initiate DNApackaging as a result of the mutation in gene 17. The protein shellsof the empty proheads expanded spontaneously (at 37 °C),resulting in mature empty heads (Fig. 2A). These were purifiedby CsCl gradient centrifugation and ion-exchange chromatog-raphy (Materials and Methods).Samples of the purified heads were frozen onto Lacey carbon

EM grids and examined with a Titan Krios electron microscope(FEI). Three different data sets were collected. An initial cryo-EM dataset included 40,000 particles recorded on film andresulted in 3D reconstruction of approximately 5.2-Å resolution.

A second cryo-EM data set was collected with the FEI Falcon IIdirect detector (pixel size of 1.38 Å) and included approximately25,000 particles that generated a map with a resolution of 3.9 Å.The third data set was acquired on a Titan Krios electron mi-croscope equipped with a K2 Summit detector (Gatan), used inthe “super-resolution”mode, resulting in a pixel size of 0.81 Å. Atotal of 19,510 isometric particles were boxed from 3,500 images.Nonreferenced 2D classification was performed with the Relionprogram (39) to select 17,954 particles. The jspr program (40)was used for generation of the initial model and refinement ofthe particle orientations, assuming icosahedral symmetry. The fi-nal 3D cryo-EM reconstruction had an overall resolution of 3.3 Å,based on the “gold-standard” Fourier shell correlation (FSC) us-ing the FSC = 0.143 criterion (41) (Fig. 2B). The map showedcontinuous density for the polypeptide chains and recognizableside-chain densities (Fig. 2C).The atomic models of the gp23*, gp24*, and Soc molecules

were built by using the COOT program (42) and refined withPhenix software (43) (SI Appendix, Table S1). The density for theHoc molecules was weak and of insufficient quality for buildingan atomic model.

Structure of the T4 Isometric Head. The T4 isometric capsid shellconsists of 120 hexameric capsomers of gp23*, 11 pentamericcapsomers of gp24*, and 1 dodecamer of gp20 (Fig. 1 B and D).During the icosahedral reconstruction, 11 pentameric verticesoccupied by gp24* were averaged with the special portal vertexoccupied by gp20. Thus, the quality of the reconstruction wasslightly diminished, especially at the vertices. However, the cryo-EM density in the vertex regions of the map was easily interpretable

Fig. 2. Cryo-EM reconstruction of the T4 isometric head. (A) EM image. (B) FSC curve. (C) Electron density of two selected regions fitted with the refinedstructure.

E8186 | www.pnas.org/cgi/doi/10.1073/pnas.1708483114 Chen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

in terms of the gp24* structure, and showed no significant influencefrom the portal protein at the special vertex.The outer and inner surfaces of the capsid were found to be

negatively charged. This implies that the dsDNA genome cannotbind easily to the inner surface of the virus, making it possible forthe DNA to be ejected from the virus while infecting a host cell.In addition, the charged outer surface can prevent aggregation ofthe virus particles. The surfaces of the gp24* pentamers are morenegatively charged than those of gp23* hexamers (Fig. 3). Thismight be to further decrease the chance of the genome gettingstuck in the vertices while being ejected.The decoration protein, Soc, forms a nearly continuous sur-

face ridge that encircles each of the gp23* hexamers except atthe interface between gp23* hexamers and gp24* pentamers (Fig.1D). Clearly, the gp23*–gp24* interface is structurally differentfrom the gp23*–gp23* interface, and is inappropriate for bindingSoc. In addition, each hexamer contains one Hoc molecule at itscenter. Thus, the stoichiometry of the isometric T4 head is720 copies of gp23*, 660 copies of Soc, 120 copies of Hoc, 55 copiesof gp24*, and 12 copies of gp20. Each icosahedral asymmetric unitconsists of 12 gp23* molecules (2 hexamers), 1 gp24* molecule, 11Soc molecules, and 2 Hoc molecules (Fig. 4).

Gp24* Structure. The gp24* structure (25) has a polypeptide foldthat is similar to the structure of the bacteriophage HK97 capsidprotein (33) (Fig. 5). Furthermore, gp24* assembles into pen-tameric capsomers, similar to the HK97 capsomers (Fig. 6 A andD). The gp24* molecule has a peripheral domain (P domain),which forms the periphery of the capsomers, and an axial domain(A domain) located near the central fivefold axes of the cap-somers (Figs. 5D and 6C). The HK97 capsid protein contains anextended loop (E loop), which is covalently crosslinked to a Pdomain of another capsid protein subunit, belonging to aneighboring capsomer. However, the T4 gp24* proteins, unlikethe HK97 capsid proteins, contain an additional insertion do-main (I domain), which is located on the outer capsid surface,

forming the characteristic surface bumps (Fig. 6A). The insertiondomain is connected to the rest of the protein structure via longlinkers (I-domain linkers), which topologically correspond to theE loop in HK97. In the gp24 crystal structure determined pre-viously, these linkers were not resolved because of their flexibility(Fig. 5C). However, the structure of these I-domain linkers iswell-resolved in the present cryo-EM map. The insertion domainis bound to a neighboring gp24* subunit in the same pentamericcapsomer, thus stabilizing the capsomer structure, a feature seenonly in the T4 capsid thus far. The linkers interact with aneighboring gp24* molecule (Fig. 6 C and E). In addition, one ofthe linkers interacts with the N-terminal region of a gp23* sub-unit from one capsomer and the P domain of a gp23* subunitfrom another capsomer (SI Appendix, Fig. S1). These non-covalent interactions stabilize the structure and occur near asimilar location where the covalent intersubunit crosslinkingoccurs in HK97 (33).The cryo-EM structure of pentameric gp24* in the T4 isometric

capsid, analyzed here, has large conformational changes comparedwith the crystal structure of the monomeric, uncleaved gp24 (Fig. 5C andD). A total of 246 of 427 residues can be aligned with an rmsdof 3.3 Å between Cα atoms. The first 10 residues of gp24 are missingin the cryo-EM structure as a result of proteolytic cleavage duringcapsid assembly. In the crystal structure, these 10 amino-terminalresidues form a helix. Residues 11–36 form an elongated N-arm inthe cryo-EM structure, whereas they form a helix in the crystalstructure. In the cryo-EM structure, the N-arm of gp24* interactswith two gp23* molecules, thereby stabilizing the capsid. In addition,a part of the N-arm (residues 26–30) is involved in a parallelβ-structure interaction with an I-domain linker of an adjacent gp24*subunit (Fig. 6E). In the mature capsid, the N terminus of gp24*(residue 11) is located on the outer surface of the capsid. However,in the prohead, the N terminus is required to be exposed to thematuration protease, which acts from within the prohead (24).In the cryo-EM structure of the pentameric gp24*, the in-

sertion domain is rotated and repositioned relative to its positionin the crystal structure of the monomeric gp24. In addition the“spine” helix (residues 194–221) in the P domain of thegp24 crystal structure is bent, but it is straight in the cryo-EMstructure of gp24* (Fig. 5C). This results in displacing theneighboring small helix (residues 181–190) and the P loop (res-idues 363–372) by approximately 7 Å.The assembly of phage heads includes a maturation step, in-

duced by a proteolytic event, resulting in an expansion of the headas it is filled with the DNA genome. The structure of gp24 beforecleavage has greatest similarity to the HK97 prohead (44) (SIAppendix, Table S2), whereas the structure of the cleaved gp24*has the greatest similarity to the HK97 mature head (33, 45) (SIAppendix, Table S2). This suggests that the crystal structure ofmonomeric gp24 and the cryo-EM structure of gp24* correspondto the immature and mature states of the T4 head, respectively.

Gp23* Structure. As anticipated, the structure of gp23* is similarto that of gp24* (rmsd of 2.6–3.0 Å; Fig. 5 E and F and SI Ap-pendix, Table S3). Like gp24*, gp23* has an insertion domain inaddition to the P and A domains of the HK97 fold (33). Theinsertion domain of gp23* is also connected via long linkerpolypeptides, which are topologically equivalent to the E-loop ofHK97 (Fig. 5E). Gp23* does not include the 65 N-terminalresidues of gp23 because these residues were proteolyticallycleaved off during capsid assembly. Unlike HK97 and gp24*, theN-terminal region of gp23* contains a compact region, which wehere refer to as the “N-fist” (residues 66–93; Fig. 5E), followedby a long N-arm (residues 94–110), similar to HK97. The Nterminus of gp23* is located on the outer surface of the gp23*shell. However, in the immature prohead the gp23 cleavage siteis located inside the prohead, where it is accessible to the pro-head protease. Hence, presumably, the N-terminal region of

Fig. 3. The molecular surfaces of gp23 and gp24 colored according toelectrostatic potential. The color ranges from red, corresponding to a po-tential of −10 kT/e−, to blue, corresponding to a potential of +10 kT/e−.

Chen et al. PNAS | Published online September 11, 2017 | E8187

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

gp23 adopts a different conformation in the prohead than thatobserved for gp23* in the mature capsid.The icosahedral asymmetric unit of the cryo-EM structure

contains two gp23* hexamers giving a total of 12 independent gp23*subunits (Fig. 4). The rmsd between corresponding Cα atoms be-tween pairs of subunits varies from 0.2 to 1.7 Å (SI Appendix, Fig.S2 and Table S4). The gp23* subunit A (Fig. 4) facing the gp24*pentamer has the greatest difference to all other subunits. Thesubunits within the hexamer closest to the pentamer (subunits A–F)have greater structural variability than the subunits within the otherhexamer of the asymmetric unit (subunits G–L). Furthermore, thehexamer closest to the pentamer has greater similarity between thesubunits related by twofold symmetry than those related by sixfoldsymmetry (Fig. 4 and SI Appendix, Table S4).The differences among the gp23* subunits are in the N-terminal

region (residues 66–110), the region that includes the insertiondomain and the I-domain linkers (residues 129–252), and in thesmall helix within the P domain (residues 272–282; SI Appendix,Fig. S2). As these regions are involved in the intersubunit inter-actions (Fig. 6D), these differences are probably necessary for thesubunits to adapt to slightly different environments in the capsid.Each gp23* subunit interacts with four other subunits within

its own hexamer and five subunits in neighboring capsomers.(Fig. 6D and SI Appendix, Fig. S3). Like in the case of gp24*, theinsertion domain and the I-domain linkers make extensive con-tacts with a neighboring subunit from the same capsomer, thusstabilizing the capsomer structure. The I-domain linkers alsointeract with the N-fists of the gp23* subunits from other cap-somers. In addition, one of the I-domain linkers interacts with

the P-loop of another subunit located in a neighboring capsomer.Similar structural motifs (P-loop and E-loop) in HK97 are in-volved in the chainmail covalent cross-linking of subunits (33).The N-arm of gp23* interacts with an I-domain linker of a

neighboring subunit belonging to the same capsomer. A part of theN-arm (residues 98–100) is involved in a parallel β-structure in-teraction with an I-domain linker of an adjacent gp23* subunit (Fig.6F). The N-fist of each gp23* interacts with four different subunits,two from the same capsomer and two from another capsomer (SIAppendix, Fig. S3). In addition, the intercapsomer interactions arestabilized by attractive electrostatic forces between the P domains ofgp23* subunits located in neighboring capsomers. The negativelycharged small helix from one P domain (residues 272–282) interactswith a positively charged β-sheet region of another P domain. Inparticular, Glu273 from the small P-domain helix interacts withLys474 and Arg476 from another gp23* subunit. Electrostatic inter-actions between similar regions were also observed in the matureand immature heads of HK97 (33, 46). Collectively, a large networkof intra- and intercapsomer interactions described here produce ahighly stable capsid structure.

Soc Structure. The 9-kDa Soc protein attaches to the outer sur-face of the T4 head during the final stage of maturation (27).The cryo-EM structure shows that each Soc molecule interactswith three gp23* subunits, which belong to two different hex-americ capsomers (Fig. 4B and SI Appendix, Fig. S4). A Socbinding site is made by the N-fist of one gp23* subunit, anI-domain linker from a second subunit belonging to the samecapsomer, and an I-domain linker from another subunit belonging

Fig. 4. Organization of the T4 isometric capsid. (A) Schematic representation of a part of the surface of the T4 head. Three asymmetric units are highlightedin blue, green, and brown. The gp24* vertex pentameric capsomers are shown as pentagons in magenta. Gp23* capsomers are shown as hexagons. The redtriangles represent the icosahedral threefold axes. The angles between adjacent capsomeres are shown in degrees on a yellow background. (B) An en-largement of the blue asymmetric unit shown in A. The asymmetric unit consists of 12 gp23* molecules (2 hexamers), 1 gp24* molecule, 11 Soc molecules(red), and 2 Hoc molecules (shown diagrammatically as a dotted circle).

E8188 | www.pnas.org/cgi/doi/10.1073/pnas.1708483114 Chen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

to a different capsomer. Furthermore, the intercapsomer inter-actions are stabilized by trimeric interactions of Soc molecules atthe quasi-threefold axes that relate three adjacent gp23* cap-somers. Reinforcement by 200 such Soc triple clamps and 60 Socmonomers near the pentameric vertices that form an external,nearly continuous, cage (Fig. 1D) allows the T4 phage to surviveunder extreme conditions such as pH 11.In the isometric capsid studied here, Soc binding sites were

not fully occupied, and the occupancies of different sites (as

determined in the structure refinement procedure) varied fromapproximately 0.6 to 0.4 (SI Appendix, Table S5). The differencein the Soc binding was the result of small conformationalchanges in the gp23* I-domain linkers, which create part of theSoc binding sites. The Soc occupancy correlated with the angles(Fig. 4A) between the normals of the planes of adjacent hex-americ capsomers (Fig. 7 and SI Appendix, Table S5). Thesmaller the angle (i.e., the more nearly the neighboring hexamersare coplanar), the lower is the occupancy of the corresponding

Fig. 5. Comparison of the T4 gp23 and gp24 monomer structures with HK97 capsid protein. (A and B) The immature and mature HK97 capsid structures.(C) The crystal structure of gp24. (D) The cryo-EM structure of gp24* reported here. (E) The cryo-EM structure of gp23* reported here. (F) The superposition ofgp23* and gp24*.

Chen et al. PNAS | Published online September 11, 2017 | E8189

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

Soc molecules. Thus, Soc provides the greatest reinforcementnear the vertices, where there is probably the greatest strain inthe gp23* hexagonal lattice. Earlier work had also shown that thestability of Soc binding to the gp23* lattice varied with localcurvature (15, 47).

Hoc Density. The crystal structure of a Hoc molecule had beendetermined previously (28), although the structure was missing thecapsid binding C-terminal domain 4 (48). The first three Ig-likedomains form an elongated but not entirely straight structure. Asthere are six possible orientations about the hexamer axis withwhich a Hoc monomer can bind to the center of a gp23* hexamer,the cryo-EM map represents an averaged structure of the Hocmolecule. Moreover, the elongated Hoc molecule is likely to beflexible. As a result, the Hoc density is difficult to interpret.The asymmetric unit of the isometric head contains two Hoc

densities, located at the centers of the two independent gp23*hexamers (Fig. 4). The Hoc density at the center of the hexamerclose to the icosahedral threefold axis was less interpretablecompared with the Hoc density in the hexamer located close tothe fivefold axis. Because the gp23* hexamer located near thefivefold vertices was less sixfold-symmetric compared with theother hexamer, there are probably preferred orientations forHoc binding to this hexamer, which in turn results in the partiallyinterpretable Hoc density. The interpretable Hoc features werean α-helix and two short β-strands, consistent with the secondarystructure prediction for the C-terminal domain of Hoc (28).

Mutations That Determine Length of the Capsid. A variety of mu-tations in gp23 can alter the WT prolate structure of the head toisometric and giant heads (15, 36–38). The T4 isometric head usedin the present study had an A275T mutation in gp23, near thequasi-threefold axis. This mutation is located in the negativelycharged small helix of the gp23* P domain, which is involved inelectrostatic interactions with the P domain from another cap-somer (SI Appendix, Fig. S5). Although the side chain of thisresidue does not interact directly with the neighboring capsomer,the mutation probably modifies the position of the interactinghelix, which in turn affects the intercapsomer interaction andangle. Another group of mutations, similar to A275T, that changethe length of the head, are A268V, D285V, D287V, and D287N(SI Appendix, Figs. S5 and S6). All these are located in a relativelysmall region of gp23*, close to the short P-domain helix, andtherefore may affect intercapsomer interactions.Mutations A66D, A94V, and G97S also change the prolate

capsid length. All these are located in the N-terminal region ofgp23. Residue 66 is the N-terminal residue of gp23*, whereasresidues 94 and 97 are located in the N-arm of gp23* (SI Ap-pendix, Fig. S6). In the prohead, the N-terminal region of gp23 islocated inside the capsid and probably interacts with the innerscaffolding core. Therefore, the gp23–core interactions play arole in the head length determination.The third cluster of mutations that change the capsid length

are T457A, R460C, and G461S. These sites are also located near thequasi-threefold axes, relating adjacent capsomers (SI Appendix, Figs.S4 and S5). In this quasi-threefold region, P-loops belonging to the

Fig. 6. Gp24 pentamer and gp23 hexamer struc-tures. (A and B) Side views of the gp24 pentamer andthe gp23 hexamer. The I domains of each subunit arelocated on the outer capsid surface forming thecharacteristic surface bumps. (C and D) Top views ofgp24 pentamer and gp23 hexamer (view from theoutside of the capsid shell). (E) Magnification of Cshows that a part of the N-arm (residues 26–30) isinvolved in a parallel β-structure interaction with anI-domain linker of an adjacent gp24* subunit.(F) Magnification of D shows that a part of the N-arm(residues 98–100) is involved in a parallel β-structureinteraction with an I-domain linker of an adjacentgp23* subunit.

E8190 | www.pnas.org/cgi/doi/10.1073/pnas.1708483114 Chen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

surrounding quasi-threefold–related capsomers interact with eachother. In the T4 prohead, these mutations may affect the P-loopconformations and consequently intercapsomer interactions.In summary, most of the gp23 mutations that change the

capsid length are located in the P domain of gp23 and likelyaffect intercapsomer binding in the prohead. However, somemutation sites are located in the N-terminal region of gp23 andlikely affect gp23 binding to the scaffolding core.

Gp24-Bypass Mutations. “Gp24-bypass mutations” are mutationsin gene 23 that allow gp23 to substitute for gp24 at the pentamericvertices when gp24 is absent (15, 49). Such mutations are N381S,A387T, and K420R in gp23* (49). All these sites are located in theA domain of gp23* that forms the core of the capsomer (SI Ap-pendix, Fig. S6). Residues 381 and 387 are located within a shortα-helix, whereas residue 420 is located in a loop region. Superpo-sition of the gp23* structure onto the structure of the pentamericgp24* vertex suggests that these residues are located in the inter-subunit interface. Therefore, these mutations affect interactionsbetween subunits within the pentameric vertices.

Placement of the Fivefold Vertices. Caspar and Klug (50) hadpredicted that, for each of the covalently identical subunits thatcompose the surface of a virus to have identical environments, itwould require that the subunits are organized into an hexagonalarray. An icosahedron is formed by substituting a pentagon ofsubunits for a hexagon of subunits at regular positions. Thiswould then allow each subunit to have at least a quasi-equivalentenvironment. The total size of the assembly is determined bywhere the pentamers replace hexamers. This prediction has beenfound to be true in a large variety of viruses with T numbersvarying from 1 for the smallest viruses such a parvoviruses (51)and the ΦX174 bacteriophage (52) to very large dsDNA viruseswith T numbers of 169 [PBCV-1 (53–55)] and 972 ≤ T ≤ 1,200[Mimivirus (56)]. Here we have determined the structure of avirus with a T=13 lattice, which makes it possible to examine howthe assembly process has introduced pentamers at specific po-sitions in the hexagonal lattice.If the subunits that make a hexamer have identical contacts

with their neighbors, they must be related by an exact sixfold axis.In turn, this implies that the centers of mass of each subunitwithin a hexamer must be coplanar. Similarly, the subunits thatcompose a pentamer must be coplanar. The T4 head structurewas determined assuming icosahedral symmetry. Thus, the sub-units in the pentamer have to be coplanar as they coincide withthe icosahedral fivefold axes. The deviation of the centers of

mass from their mean plane through the two independent hex-amers per icosahedral asymmetric unit of the T4 isometric headare given in SI Appendix, Table S6. No distance was greater than0.8 Å. Furthermore, superposition of one hexamer onto theother showed an rmsd of 1.5 Å between equivalent Cα atoms,with only 1% of Cα atoms deviating by more than 4 Å, which isprobably not significant considering the 3.3-Å resolution of thecryo-EM map. If the hexamers and pentamers are planar, theangle between hexamers or between pentamers and hexamerswill be determined by the T number. For instance, the anglebetween adjacent fivefold axes in an icosahedron is 63.44°. Thus,for a virus with T=1 symmetry, the angle between the normals oftwo adjacent pentamers has to be 63.44°. Hence, the amino acidsalong the periphery of a pentamer of a T=1 virus must bind twopentamers together so as to produce an angle of 63.44° betweenthe pentamers. Larger T numbers require the introduction ofhexamers into the array forming the capsid. Thus, the curvaturebetween pentamers will be spread over the angular differencebetween adjacent pentamers and hexamers and between adja-cent hexamers. For instance, in the present case of a T=13 capsid(Fig. 4A), the angular difference between the “blue pentamer”and the “green pentamer” is given roughly by the sum of theangles between the blue pentamer and the blue ABCDEF hex-amer (29°), the blue ABCDEF hexamer and blue GHIJKLhexamers (6°), the blue GHIJKL hexamer and the green ABCDEFhexamer (18°; a sum of 53° so far), and also possibly between thegreen ABCDEF hexamer and the green pentamer (29°), whichwould give a total of 82°. This is not exactly the 63.44° required foricosahedral symmetry because the individual rotations are not ex-actly about a common axis of rotation and because the individualhexamers and pentamers do not necessarily have exact five- orsixfold symmetry.On strictly geometric grounds, the angles between the normal

to a pentamer and each of the normals to the surrounding hexamershave to be 37.62° to create a pentagonal hole in a hexagonal array.Furthermore, also on strictly geometric grounds, the angle betweenthe normals of the adjacent hexamers surrounding a pentamer hasto be 41.81°. These angles are fixed for any icosahedral assembly ofpentamers and hexamers. The observed angles were 29° and 33°,respectively (Fig. 4A), for the isometric T4 assembly. The deviationof these angles from the theoretical values of 37.62° and 41.81°,respectively, must be the result of a lack of exact planarity and alack of exact rotational symmetry.Any mutation that alters the T number must have the property

to change the angles between capsomers and therefore mustoccur at the periphery of the gp23* hexamers (SI Appendix, Fig.S6). Furthermore, mutations that change the T number would bemore easily accommodated if the mutated amino acid were hy-drophobic and therefore independent of the exact hydrophobicenvironment. In contrast, mutations of polar residues wouldrequire more accurate positioning with respect to the sur-rounding polar environment. Thus, for an amino acid to achievea variety of similar but different contacts, as would be the case ifthe mutation changed the T number, the mutated amino acidwould more likely have to be hydrophobic rather than polar.Hence, mutations that increase the variety of polymorphic par-ticles might tend to increase the hydrophobicity of an amino acidat the interface between capsomers. This is roughly consistentwith the observed variety of T4 head length-varying mutationsdescribed here.

ConclusionsWe report the 3.3-Å resolution structure of the ∼50-MDa isometricbacteriophage T4 capsid that has a diameter of approximately 860 Å.This is one of the largest structures determined to date to near-atomic resolution. The major capsid protein, gp23*, has a poly-peptide fold similar to the phage HK97 capsid protein. However,unlike the HK97 capsid that is stabilized by covalent crosslinking of

Fig. 7. Correlation between the intercapsomer angles and Soc occupancy.

Chen et al. PNAS | Published online September 11, 2017 | E8191

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

subunits (33), the T4 capsid shows intricate noncovalent interactionsamong the subunits that, together, generate a very stable virus capsid.The T4 major capsid protein subunits are “glued” together by

the I domains that make a network of interactions with the A and Pdomains of other gp23* subunits in the same hexameric capsomer.The I-domain is flanked by long linkers, which are topologicallyequivalent to the elongated “E” loop of the HK97 fold. TheI-domain linkers are involved in extensive intra- and intercapsomerinteractions that stabilize the capsid. The intercapsomer interac-tions are reinforced by electrostatic forces between the P domainsof gp23* subunits from adjacent capsomers. Furthermore, theN-terminal regions of gp23* interact extensively with four neigh-boring subunits belonging to two different capsomers. These sta-bilizing interactions allow the capsid to withstand the internalpressure from the tightly packed DNA.The capsid is further stabilized by Soc molecules, which form

an external cage of 200 trimers, each clamping three adjacentcapsomers. The Soc protein has the highest affinity for theirbinding sites located near pentameric vertices where there is thegreatest strain on the capsids’ integrity, in part because of thebigger angle between the planes of adjacent capsomers.Mutations that change the capsid length are mostly located in

the capsomer’s periphery and therefore affect the angle betweenthe planes of adjacent capsomers. In turn, this affects the posi-tions where pentameric holes are created within the hexagonalarray of hexamers, resulting in different triangulation numbers.Many of the mutations result in greater hydrophobicity of thebinding interface, thereby probably imparting flexibility to sup-port various environments at the intercapsomer interfaces asrequired by different T numbers.The architecture of T4 capsid illustrates how an extremely stable

virus capsid can be constructed, yet with considerable plasticity toalter the size of the capsid. Thus, relatively little change in thesubunit structure can assemble stable structures of different size.The same principles will be relevant when considering the assemblyof very small viruses such as ΦX174 (52) and very large viruses suchas adenovirus (57–59), PBCV1 (55), and even Mimivirus (56).

Materials and MethodsPreparation of the Isometric Mutant Phage. Stocks of 17amNG178.23pt21–34cmutant phage were prepared on the amber suppressor E. coli strain B40. Con-struction of this mutant was described previously (31). Isometric phage can ac-commodate only >65% of phage T4 genome, hence it cannot productivelyinfect E. coli and produce a plaque. Therefore, there is strong selection pressureto accumulate WT revertants. To minimize this, the 17amNG178.23pt21–34cphage stocks were prepared starting from a minute plaque produced on theamber suppressor E. coli strain B40. Large stock was then amplified by sub-culturing the minute plaque no more than twice to minimize the accumulationof revertants. Consequently, the plaque-forming titer of the phage stock is quitelow, 1–3 × 1010 pfu/mL, but the number of virus particles as determined by SDS/PAGE is approximately 10 times higher. Heads produced from such stockscontained ∼80% isometric empty capsid particles.

Purification of Isometric Capsids. E. coli P301 (suppressor-minus) cells weregrown at 37 °C on 1:1 mixture of Luria broth and M9 minimal mediumsupplemented with amino acids, infected with 17amNG178.23pt21–34cphage at a multiplicity of infection (MOI) of 2.5, and superinfected at the sameMOI 7 min later. Infection was allowed to continue for another 20 min. The cellswere pelleted by centrifugation at 7,000 rpm (4,300 × g) for 12 min. The pelletwas then resuspended in a buffer containing 50 mM Tris·HCl, pH 7.5, 5 mMMgCl2, 200 mM NaCl, and 150 mM imidazole. These conditions produced ∼90%expanded empty proheads, of which ∼80% were isometric. However, most ofthese heads were tailed by nonspecific binding of neck and tail proteins. Highconcentration of salt and imidazole in the buffer dissociated the tails andgenerated tailless empty proheads.

The crude extract of tailless isometric capsids prepared as described earlierwas treated with DNaseI (20 μg/mL) to digest the phage DNA. Several drops ofchloroform were added to lyse the cells, and the sample was incubated at37 °C for 30 min. The lysate was then centrifuged at 7,000 rpm (4,300 × g) for10 min, and the pellet containing cell debris was discarded. The supernatantcontaining the heads was centrifuged at 17,000 rpm (34,500 × g) for 45 min.

The pellet was resuspended in a buffer containing 50 mM Tris·HCl, pH 7.5, and5 mM MgCl2, and layered on a 5-mL cesium chloride gradient. The gradientswere centrifuged at 40,000 rpm for 1 h in a SWTi55 rotor. The proheadsbanded near approximately half the length of the gradient tube. The topportion of the band, which predominantly contained isometric proheads, wascollected and dialyzed overnight against 10 mM Tris·HCl, pH 7.5, 50 mM NaCl,and 5 mM MgCl2. The proheads were then purified by chromatography byusing two ion-exchange columns, first on a DEAE-Sephacel column and secondon a Mon-Q FPLC column. In each case, a linear 50–300-mM NaCl gradient wasused to separate the expanded (i.e., mature) isometric capsids from anyunexpanded capsids and contaminating DNA. The peak fractions containingthe expanded particles were concentrated by Amicon membrane filter (30Kcutoff) to approximately 1010 particles per microliter and stored at −80 °C.

Preparation of Isometric Heads for Cryo-EM. Purified T4 isometric head samplewas frozen by using a Gatan CP3 cryo-plunger. A 2.5-μL sample was appliedonto the glow-discharged Lacey carbon grid with 3-nm-thick supportingcarbon film, with a blotting time of 7 s with humidity maintained at greaterthan 85%. The grids were stored in liquid nitrogen for future use.

Cryo-EM Imaging, Data Processing, and Resolution Assessment. An initial cryo-EM dataset was collected by using films installed on a Titan Krios microscope(FEI) at Purdue University. Approximately 40,000 particles were boxed for re-construction from 1,512 films with pixel size of 1.1 Å per pixel. A reconstructionwas calculated by using the program EMAN (60, 61), obtaining a resolution of5.2 Å. The backbone structures of gp23* and gp24* were built with the guideof the previous gp24 crystal structure and gp23 predicted model.

A second cryo-EM data set was collected with a Falcon II direct detectorinstalled on a Titan Krios microscope (FEI) at Sanford-Burnham-PrebysMedical Discovery Institute. Approximately 4,000 images were collectedwith a pixel size of 1.38 Å per pixel. Approximately 25,000 particles wereboxed. Each image was composed of seven frames, with a total dose rateof approximately five electrons per square Ångström. The resulting re-construction was calculated by using jspr (40) and had a resolution of 3.9 Å.

The final cryo-EM data set was collected by using the Purdue UniversityTitan Krios microscope equipped with a Gatan K2 Summit detector. Ap-proximately 3,500 images were collected by using a magnification of180,000×, with pixel size of 0.81 Å at superresolution mode. Each image wascomposed of 45 frames. Each frame had an exposure time of 0.2 s and a doserate of approximately three electrons per Ångström.

Motion correction was applied to each image (62). Averaged images wereproduced from all frames in the stack. The contrast transfer function correc-tion was applied to each motion-corrected image. These images were used forboxing the particles. A total of 19,510 particles were boxed. A reference-free,2D classification program implanted in the RELION software (39) was used toclassify the boxed particles by a reciprocal space maximum-likelihood method.The 2D classification selected 17,954 particles for further data processing. Theprogram jspr (40) was used for producing an initial model and further 3Drefinement. The jspr program was also used to correct the anisotropic mag-nification of images produced by the microscope (63). To refine the 3D densitymap, the same particles were reboxed from the averaged images by using onlyframes 3–22 while excluding the initial 2 frames with large movements andthe last 23 frames with greater radiation damage. The program jspr (40) wasused for the refinement. The dataset of reboxed particles was split into twohalves before the start of the refinement. The low-resolution model was thenrefined in each half dataset independently. The refinement stopped whenthere was no further improvement of the FSC. The FSC was calculated from thetwo independent reconstructions. The resolution of the reconstruction wasdetermined by the point at which the FSC was 0.143. Both datasets werecombined for calculation of the final map. A temperature factor of −100 Å2

and a low-pass filter to 3.3 Å was applied for calculation of the final 3D map.

Atomic Model Building and Refinement. The map showed continuous densityfor the main chain and distinct side-chain densities for bulky amino acidresidues. Based on this map, the atomic models of the gp23*, gp24*, and Socproteins within one asymmetric unit were built by using the COOT graphicalprogram (42). The structure was refined in real space against the experi-mental electron density by using the phenix.real_space_refine program (64)from Phenix software (43) (SI Appendix, Table S1). The occupancies of the 11Soc molecules located in the asymmetric unit were refined in reciprocalspace by using the phenix.refine program (43).

Figures were prepared by using the programs Chimera (65) and Pymol(www.pymol.org).

E8192 | www.pnas.org/cgi/doi/10.1073/pnas.1708483114 Chen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0

ACKNOWLEDGMENTS. We thank Valorie Bowman for help with the datacollection at Purdue University, Mark Swift and Niels Volkmann for help with datacollection at Sanford-Burnham-Prebys Medical Discovery Institute, and Wei-ChunTang for help with DNA sequencing of T4 phage mutants at The Catholic

University of America. This work was supported by National Institutes of Health(NIH) Grant AI081726 (to V.B.R. and M.G.R.). NIH Grants S10 OD012372 and P01GM098412-S1 (to D.H.) funded the purchase of the Titan Krios and Falcon II directdetection imaging device at Sanford-Burnham-Prebys Medical Discovery Institute.

1. Hatfull GF, Hendrix RW (2011) Bacteriophages and their genomes. Curr Opin Virol1:298–303.

2. Hendrix RW (2003) Bacteriophage genomics. Curr Opin Microbiol 6:506–511.3. Evilevitch A, Lavelle L, Knobler CM, Raspaud E, Gelbart WM (2003) Osmotic pressure

inhibition of DNA ejection from phage. Proc Natl Acad Sci USA 100:9292–9295.4. Tzlil S, Kindt JT, Gelbart WM, Ben-Shaul A (2003) Forces and pressures in DNA pack-

aging and release from viral capsids. Biophys J 84:1616–1627.5. Leiman PG, Kanamaru S, Mesyanzhinov VV, Arisaka F, Rossmann MG (2003) Structure

and morphogenesis of bacteriophage T4. Cell Mol Life Sci 60:2356–2370.6. Yap ML, Rossmann MG (2014) Structure and function of bacteriophage T4. Future

Microbiol 9:1319–1327.7. Liu S, et al. (2014) A viral packaging motor varies its DNA rotation and step size to

preserve subunit coordination as the capsid fills. Cell 157:702–713.8. Berndsen ZT, Keller N, Smith DE (2015) Continuous allosteric regulation of a viral

packaging motor by a sensor that detects the density and conformation of packagedDNA. Biophys J 108:315–324.

9. Sun S, et al. (2008) The structure of the phage T4 DNA packaging motor suggests amechanism dependent on electrostatic forces. Cell 135:1251–1262.

10. Rao VB, Feiss M (2015) Mechanisms of DNA packaging by large double-stranded DNAviruses. Annu Rev Virol 2:351–378.

11. Sathaliyawala T, et al. (2006) Assembly of human immunodeficiency virus (HIV) an-tigens on bacteriophage T4: A novel in vitro approach to construct multicomponentHIV vaccines. J Virol 80:7688–7698.

12. Shivachandra SB, et al. (2007) Multicomponent anthrax toxin display and deliveryusing bacteriophage T4. Vaccine 25:1225–1235.

13. Tao P, Mahalingam M, Rao VB (2016) Highly effective soluble and bacteriophageT4 nanoparticle plague vaccines against Yersinia pestis. Methods Mol Biol 1403:499–518.

14. Tao P, et al. (2013) In vitro and in vivo delivery of genes and proteins using thebacteriophage T4 DNA packaging machine. Proc Natl Acad Sci USA 110:5846–5851.

15. Black LW, Showe MK, Steven AC (1994) Morphogenesis of the T4 head. MolecularBiology of Bacteriophage T4, ed Karam JD (Am Soc Microbiol, Washington, DC), pp218–258.

16. Fokine A, et al. (2004) Molecular architecture of the prolate head of bacteriophageT4. Proc Natl Acad Sci USA 101:6003–6008.

17. Kostyuchenko VA, et al. (2005) The tail structure of bacteriophage T4 and its mech-anism of contraction. Nat Struct Mol Biol 12:810–813.

18. Leiman PG, et al. (2010) Morphogenesis of the T4 tail and tail fibers. Virol J 7:355.19. Kostyuchenko VA, et al. (2003) Three-dimensional structure of bacteriophage T4

baseplate. Nat Struct Biol 10:688–693.20. Taylor NM, et al. (2016) Structure of the T4 baseplate and its function in triggering

sheath contraction. Nature 533:346–352.21. Yap ML, et al. (2016) Role of bacteriophage T4 baseplate in regulating assembly and

infection. Proc Natl Acad Sci USA 113:2654–2659.22. Leiman PG, Chipman PR, Kostyuchenko VA, Mesyanzhinov VV, Rossmann MG (2004)

Three-dimensional rearrangement of proteins in the tail of bacteriophage T4 on in-fection of its host. Cell 118:419–429.

23. Sun L, et al. (2015) Cryo-EM structure of the bacteriophage T4 portal protein assemblyat near-atomic resolution. Nat Commun 6:7548.

24. Fokine A, Rossmann MG (2016) Common evolutionary origin of procapsid proteases,phage tail tubes, and tubes of bacterial type VI secretion systems. Structure 24:1928–1935.

25. Fokine A, et al. (2005) Structural and functional similarities between the capsid pro-teins of bacteriophages T4 and HK97 point to a common ancestry. Proc Natl Acad SciUSA 102:7163–7168.

26. Rao VB, Black LW (1985) DNA packaging of bacteriophage T4 proheads in vitro. Ev-idence that prohead expansion is not coupled to DNA packaging. J Mol Biol 185:565–578.

27. Ishii T, Yanagida M (1977) The two dispensable structural proteins (soc and hoc) of theT4 phage capsid; their purification and properties, isolation and characterization ofthe defective mutants, and their binding with the defective heads in vitro. J Mol Biol109:487–514.

28. Fokine A, et al. (2011) Structure of the three N-terminal immunoglobulin domains ofthe highly immunogenic outer capsid protein from a T4-like bacteriophage. J Virol 85:8141–8148.

29. Qin L, Fokine A, O’Donnell E, Rao VB, Rossmann MG (2010) Structure of the smallouter capsid protein, Soc: A clamp for stabilizing capsids of T4-like phages. J Mol Biol395:728–741.

30. Fokine A, et al. (2013) The molecular architecture of the bacteriophage T4 neck. J MolBiol 425:1731–1744.

31. Iwasaki K, et al. (2000) Molecular architecture of bacteriophage T4 capsid: Vertexstructure and bimodal binding of the stabilizing accessory protein, Soc. Virology 271:321–333.

32. Olson NH, Gingery M, Eiserling FA, Baker TS (2001) The structure of isometric capsidsof bacteriophage T4. Virology 279:385–391.

33. Wikoff WR, et al. (2000) Topologically linked protein rings in the bacteriophageHK97 capsid. Science 289:2129–2133.

34. Fokine A, Rossmann MG (2014) Molecular architecture of tailed double-stranded DNAphages. Bacteriophage 4:e28281.

35. Baker ML, Jiang W, Rixon FJ, Chiu W (2005) Common ancestry of herpesviruses andtailed DNA bacteriophages. J Virol 79:14967–14970.

36. Doermann AH, Pao A, Jackson P (1987) Genetic control of capsid length in bacte-riophage T4: Clustering of ptg mutations in gene 23. J Virol 61:2823–2827.

37. Lane T, Eiserling F (1990) Genetic control of capsid length in bacteriophage T4. VII. Amodel of length regulation based on DNA size. J Struct Biol 104:9–23.

38. Mooney DT, Stockard J, Parker ML, Doermann AH (1987) Genetic control of capsidlength in bacteriophage T4: DNA sequence analysis of petite and petite/giant mu-tants. J Virol 61:2828–2834.

39. Scheres SH (2012) RELION: Implementation of a Bayesian approach to cryo-EMstructure determination. J Struct Biol 180:519–530.

40. Guo F, Jiang W (2014) Single particle cryo-electron microscopy and 3-D reconstructionof viruses. Methods Mol Biol 1117:401–443.

41. Henderson R, et al. (2012) Outcome of the first electron microscopy validation taskforce meeting. Structure 20:205–214.

42. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.Acta Crystallogr D Biol Crystallogr 66:486–501.

43. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D Biol Crystallogr 66:213–221.

44. Gertsman I, et al. (2009) An unexpected twist in viral capsid maturation. Nature 458:646–650.

45. Helgstrand C, et al. (2003) The refined structure of a protein catenane: TheHK97 bacteriophage capsid at 3.44 A resolution. J Mol Biol 334:885–899.

46. Gertsman I, Fu CY, Huang R, Komives EA, Johnson JE (2010) Critical salt bridges guidecapsid assembly, stability, and maturation behavior in bacteriophage HK97. Mol CellProteomics 9:1752–1763.

47. Aebi U, et al. (1977) Capsid fine structure of T-even bacteriophages. Binding andlocalization of two dispensable capsid proteins into the P23* surface lattice. J Mol Biol110:687–698.

48. Sathaliyawala T, et al. (2010) Functional analysis of the highly antigenic outer capsidprotein, Hoc, a virus decoration protein from T4-like bacteriophages. Mol Microbiol77:444–455.

49. Johnson K, Condie B, Mooney DT, Doermann AH (1992) Mutations that eliminate therequirement for the vertex protein in bacteriophage T4 capsid assembly. J Mol Biol224:601–611.

50. Caspar DL, Klug A (1962) Physical principles in the construction of regular viruses. ColdSpring Harb Symp Quant Biol 27:1–24.

51. Tsao J, et al. (1991) The three-dimensional structure of canine parvovirus and itsfunctional implications. Science 251:1456–1464.

52. McKenna R, et al. (1992) Atomic structure of single-stranded DNA bacteriophage phiX174 and its functional implications. Nature 355:137–143.

53. Yan X, Olson NH, Van Etten JL, Baker TS (1998) Cryo-electron microscopy and image re-construction of PBCV-1, an algal virus with T=169 lattice symmetry. Electron Microscopy,1998: Proceedings of the 14th International Congress on Electron Microscopy, Cancún(Mexico), eds Calderón Benavides HA, Yacamán MJ (Institute of Physics, Bristol, UK), Vol4, pp 775–776.

54. Nandhagopal N, et al. (2002) The structure and evolution of the major capsid proteinof a large, lipid-containing DNA virus. Proc Natl Acad Sci USA 99:14758–14763.

55. Zhang X, et al. (2011) Three-dimensional structure and function of the Parameciumbursaria chlorella virus capsid. Proc Natl Acad Sci USA 108:14837–14842.

56. Xiao C, et al. (2009) Structural studies of the giant mimivirus. PLoS Biol 7:e92.57. Liu H, et al. (2010) Atomic structure of human adenovirus by cryo-EM reveals inter-

actions among protein networks. Science 329:1038–1043.58. Reddy VS, Natchiar SK, Stewart PL, Nemerow GR (2010) Crystal structure of human

adenovirus at 3.5 A resolution. Science 329:1071–1075.59. Rux JJ, Burnett RM (2004) Adenovirus structure. Hum Gene Ther 15:1167–1176.60. Ludtke SJ (2010) 3-D structures of macromolecules using single-particle analysis in

EMAN. Methods Mol Biol 673:157–173.61. Ludtke SJ, Baldwin PR, Chiu W (1999) EMAN: Semiautomated software for high-

resolution single-particle reconstructions. J Struct Biol 128:82–97.62. Li X, et al. (2013) Electron counting and beam-induced motion correction enable near-

atomic-resolution single-particle cryo-EM. Nat Methods 10:584–590.63. Yu G, et al. (2016) An algorithm for estimation and correction of anisotropic mag-

nification distortion of cryo-EM images without need of pre-calibration. J Struct Biol195:207–215.

64. Afonine PV, Headd JJ, Terwilliger TC, Adams PD (2013) New tool: phenix.real_space_refine. Computational Crystallography Newsletter 4:43–44.

65. Pettersen EF, et al. (2004) UCSF Chimera–a visualization system for exploratory re-search and analysis. J Comput Chem 25:1605–1612.

Chen et al. PNAS | Published online September 11, 2017 | E8193

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

30,

202

0