Structure of a headful DNA-packaging bacterial virus at2.9 Å resolution by electron cryo-microscopyHaiyan Zhaoa,1, Kunpeng Lib,1, Anna Y. Lynna, Keith E. Arona, Guimei Yub, Wen Jiangb,2, and Liang Tanga,2

aDepartment of Molecular Biosciences, University of Kansas, Lawrence, KS 66045; and bDepartment of Biological Sciences, Purdue University, WestLafayette, IN 47907

Edited by Eva Nogales, University of California, Berkeley, CA, and approved February 24, 2017 (received for review September 8, 2016)

The enormous prevalence of tailed DNA bacteriophages on thisplanet is enabled by highly efficient self-assembly of hundreds ofprotein subunits into highly stable capsids. These capsids can standwith an internal pressure as high as ∼50 atmospheres as a result ofthe phage DNA-packaging process. Here we report the completeatomic model of the headful DNA-packaging bacteriophage Sf6 at2.9 Å resolution determined by electron cryo-microscopy. The struc-ture reveals the DNA-inflated, tensed state of a robust protein shellassembled via noncovalent interactions. Remarkable global confor-mational polymorphism of capsid proteins, a network formed byextended N arms, mortise-and-tenon–like intercapsomer joints,and abundant β-sheet–like mainchain:mainchain intermolecular in-teractions, confers significant strength yet also flexibility requiredfor capsid assembly and DNA packaging. Differential formations ofthe hexon and penton are mediated by a drastic α–helix-to-β–strandstructural transition. The assembly scheme revealed here may becommon among tailed DNA phages and herpesviruses.

virus assembly | bacteriophage | DNA packaging | atomic structure |cryoEM

Tailed double-stranded DNA (dsDNA) bacteriophages are themost abundant biological entities on this planet (1, 2). The

enormous prevalence of these viruses is made possible by rapid,highly efficient and error-free self-assembly of hundreds of copiesof capsid protein subunits into phage capsids. During replication, aphage-encoded machinery called “terminase” packages the phagedsDNA genome into smaller capsid precursors called “procap-sids” in an ATP-dependent manner (3, 4), resulting in structuraltransformation, or maturation, to expanded mature capsids (5).The terminase molecules are the most powerful molecular motorsand can work against a force larger than 50 pN and package DNAat high rates, resulting in highly densely packed DNA in the capsidat a density of ∼500 g/L and conferring a pressure as high as50 atmospheres onto the capsid inner surface (6, 7), more than10 times that found in any other living system (6, 8, 9). Such aninternal pressure may have implications in injection of phageDNA into host cells upon infection (9). The phage capsids, formedvia noncovalent interactions in most cases, must be sufficientlystable to withstand such an internal pressure.The assembly of phage capsids and procapsids has been studied

intensively (5). Capsids are typically formed by a single capsid pro-tein, in some cases with additional minor proteins. The capsid pro-teins share a common fold (10), as first shown in the 3.44-Å X-raystructure of phage HK97 capsid formed by a single ∼280-residueprotein gp5 (11, 12) and later observed in other phages (10) as wellas in the floor domain of herpesvirus major capsid proteins (13, 14).A recent 3.5-Å electron cryo-microscopy (cryoEM) structure ofphage T7 shed light on the capsid assembly at the near atomic level(15). The previous HK97 structures were determined with particlesassembled from heterologously expressed capsid proteins inEscherichia coli and/or mutants, which did not contain the DNAgenome. No high-resolution structure of authentic, DNA-filled vi-rion is available, preventing understanding of the state of the capsidunder the DNA-caused high internal pressure.Bacteriophage Sf6 is a member of the Podoviridae family of

tailed dsDNA bacteriophages (16). Sf6 has an isometric icosahedral

capsid made up of a single protein gp5 of 423 residues, into whichthe dsDNA genome of ∼39 kbp is packaged using a headful DNA-packaging mechanism that is shared by many other phages such asT4, P22, and SPP1 (17). In such a mechanism, phage DNA isinserted into each capsid precursor or procapsid by the phageterminase until the head is full (thus “headful”), resulting in aterminally redundant, circularly permuted dsDNA molecule pack-aged in each capsid that is ∼6% longer than the genome (17).We have determined the capsid structure of the bacteriophage

Sf6 authentic virion at 2.9 Å resolution by cryoEM. The structurereveals atomic details of the construction of an ultra-stableprotein shell consisting of >400 copies of a single capsid pro-tein with a thickness of 12 Å at the thinnest place in its native,DNA-filled state, assembled via noncovalent interactions. Theassembly scheme may be common among tailed DNA phagesand herpesviruses and may have implications in fabrication ofstrong, but flexible, self-assembled nano-containers.

Results and DiscussionStructure Determination and Overall Structure. Phage Sf6 has anultra-stable capsid, which remains essentially intact after incubationat 60 °C for 45 min (SI Appendix, Fig. S1). Fluorescence-basedthermostability assay shows a melting temperature (Tm) at ∼92 °C.In contrast, the Tm of the Sf6 DNA-packaging terminase gp2, a470-residue protein, is ∼46 °C.The cryoEM images of frozen-hydrated phage Sf6 were recorded

with a direct electron detector operating on the superresolution

Significance

The enormous prevalence of tailed double-stranded DNA bacte-riophages on this planet is made possible by highly efficient self-assembly of usually hundreds of protein subunits into highlystable capsids. These capsids must stand with an internal pressureas a result of the phage DNA-packaging process. Here we reportthe complete atomic model of the headful DNA-packaging bac-teriophage Sf6 at 2.9 Å resolution determined by electron cryo-microscopy. The structure reveals the DNA-inflated, tensed stateof a robust yet flexible protein shell assembled via noncovalentinteractions, enabled by remarkable global conformational poly-morphism of capsid proteins, a network formed by extended Narms, and abundant β-sheet–like mainchain:mainchain intermo-lecular interactions. The assembly mechanism may be commonamong tailed DNA phages and herpesviruses.

Author contributions: H.Z., W.J., and L.T. designed research; H.Z., K.L., A.Y.L., K.E.A., G.Y.,W.J., and L.T. performed research; H.Z., W.J., and L.T. analyzed data; and H.Z., W.J., andL.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The cryoEM map has been deposited with the EMDataBank (accessioncode EMD-8314). The coordinates of the structure have been deposited with the ResearchCollaboratory for Structural Bioinformatics Protein Data Bank (entry code of 5L35).1H.Z. and K.L. contributed equally to this work.2To whom correspondence may be addressed. Email: jiang12@purdue.edu or tangl@ku.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1615025114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1615025114 PNAS | April 4, 2017 | vol. 114 | no. 14 | 3601–3606

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

counting mode (SI Appendix, Fig. S2, and Materials and Meth-ods). The structure was determined at 2.9 Å resolution (Fig. 1)with the single-particle cryoEM method, aided by correction ofanisotropic magnification distortion of the electron microscope(18). The icosahedral symmetry was imposed in the reconstruc-tion, resulting in the structure for the capsid, whereas the singletail was averaged out. The cryoEM map was of high quality,showing well-defined density for most amino-acid side chains aswell as protruding density for main-chain oxygen atoms (SI Ap-pendix, Fig. S2). This enabled automated model buildingfor >85% amino acid residues, from which a complete atomicmodel for all of the seven capsid protein subunits in the icosa-hedral asymmetric unit was constructed with additional manualrebuilding. The atomic model has excellent stereochemistry andan average real-space correlation coefficient of 0.94 (Fig. 1B andSI Appendix, Fig. S2 and Table S1).The capsid exhibits a T = 7laevo icosahedral symmetry con-

sisting of 12 pentons situated at icosahedral fivefold vertices and60 hexons, with a maximal diameter of 690 Å between two op-posite fivefold vertices (Fig. 1). There are seven capsid proteinmolecules in the icosahedral asymmetric unit. Each moleculecontains an elongated N-terminal arm (N arm; residues 1–45)(Fig. 1 and SI Appendix, Fig. S3), an extended loop (E loop;residues 46–83), a peripheral domain (P domain; residues 84–131 and 357–412), and an axial domain (A domain: residues 132–216, 351–356, and 413–423) that juxtapose around the center ofhexons and pentons. These structural elements overall match thecanonical HK97 fold in tailed dsDNA bacteriophages and

herpesviruses (SI Appendix, Fig. S4) (11, 13, 14). The Sf6 capsidprotein contains an additional 134-residue insertion domain(I domain; residues 217–350) exposed on the outer surface of thecapsid. The Sf6 P domain contains a four-stranded β-sheet in-stead of the three-stranded ones in HK97 and T7 (Fig. 1 and SIAppendix, Fig. S4). Another difference is in the region of the Adomain near the center of capsomers. In Sf6, this region exhibitsthe short helix α5 in hexons, which transitions to a β-strand inpentons (see below), whereas in HK97 this region is a loop (SIAppendix, Fig. S4) (12).

The Capsid Protein Exhibits Global Conformational Polymorphism.The seven protein subunits in the icosahedral asymmetric unitare all structurally distinct and the structural differences areglobal. The A domains are formed by a five-stranded β-sheetpacked against three helices and are superimposable among theseven molecules except for molecule G forming the penton (seebelow). The I domains alone are overall superimposable, withvariances mostly in the loop 281–290 that are involved in theinter–I-domain interactions. However, when superimposing theseven molecules according to the A domains, the I domainsundergo various shifting with a Cα atomic displacement of asmuch as 1.5 Å, indicating changes in orientations and positionsof the I domains with respect to the A domains.Significantly larger structural differences occur in the N arms,

E loops, and P domains (Fig. 1). The extended N arms undergolarge-scale swinging by as much as 35° between molecules B andG, resulting in a positional displacement of 28 Å for the terminalresidue P2 (Fig. 1). The E loops and P domains bend drasticallyby ∼15° and ∼20°, respectively, between molecules B and G. Thebending leads to significant movement at the distal ends of bothstructural elements, that is, helix α1 and the P loop (residues393–396) by ∼10 Å in the P domain and the tip of the E loop by∼9 Å. The bending in E loops and P domains is not pivoted onany single residues or atoms and cannot be explained by rigidbody movement, but instead is propagated gradually over thelength, which is enabled by a series of glycine and proline resi-dues including G84, G81, G77, and P53 in the E loop and P406,G361, and G363 in the P domain (Fig. 1E and SI Appendix, Fig.S3). The β1 and β2 in the E loop and the β21 and β25 in the Pdomain form a curled β-sheet resembling a half β-barrel, andβ2 in the E loop extends to β3 in the P domain, making the twodomains essentially a single module, which is termed the “plank”module here. The half β-barrel harbors many of the glycine andproline residues that enable the bending as described above andthus apparently serves as a structural unit that mediates thebending of the plank module. The β2 in this half β-barrel con-tains a KATG motif conserved among Sf6, phage APSE-1, andP22 (SI Appendix, Fig. S3).The extent of the plank module bending differs among the

seven molecules depending on their positions in the icosahedrallattice (Fig. 1 and Movie S1). The plank modules in molecules Band F have essentially the same conformation, but the plankmodule becomes increasingly bent in the order of molecules E,C, D, and A and culminates in molecule G. These data areconsistent with the fact that the plank module of molecule B isjuxtaposed with that of molecule F of an adjacent hexon re-lated by the icosahedral fivefold symmetry. Likewise, the plankmodule in molecule A is juxtaposed with that of molecule G,representing a transition from the hexon, where molecule A isfrom, to the penton, to which molecule G belongs, as the pentongenerates the curvature as required for formation of theicosahedral capsid.Within the context of the P domain, the four-stranded β-sheet

encompassing β3, β22, β23, and β24 exhibits twisting to variousextents among the seven molecules (SI Appendix, Fig. S5). Suchtwisting is coupled with the bending of the spine helix (residues105–131). Bending of the spine helix is propagated over thelength and culminates at residue P111, leading to significantmovement of helix α1 and the P loop. Interestingly, the sidechain of F127 in the spine helix flips over in molecules B and F

A

B

C

D

F

E

G

I domain A domain N arm

E loop

A

P domain

B

90

D

C

Plank module Plank module

35

E

90

Half -barrel

G77 G81 G84

P53

P406

G361 G363

1 2

21

25

Fig. 1. The cryoEM structure of phage Sf6. (A) The cryoEM map coloredradially from blue (outside) to light blue (inside). The hexagonal lattice of theT = 7laevo icosahedral symmetry is shown as a green stick model. The regionsfor the seven molecules in an icosahedral asymmetric unit are rainbow-coloredfrom blue for molecule A to red for molecule G. (B) Ribbon diagram of theseven molecules in the icosahedral asymmetric unit. (C ) Domains in amonomer: N arm, red; E loop, yellow; A domain, blue; P domain, green; Idomain, cyan. (Left) A view from outside of the capsid. (Right) A view 90°from that in the Left panel and approximately tangential to the capsidsurface. The E loop and P domain together are labeled as “plank module.”The spine helix is indicated with a green arrow. (D) Superimposition ofmolecules B (ribbon diagram in light blue), D (green), A (blue), and G (red),showing global conformational polymorphism, particularly differentialbending of the plank module. The view is the same as in the Right panel ofC. For clarity, molecules C, E, and F are not shown. (E) Closeup of bent plankmodules. Glycine and proline residues enabling the bending (see text) are inmagenta and labeled. For clarity, only molecules B and G are shown. β-Strandsare labeled.

3602 | www.pnas.org/cgi/doi/10.1073/pnas.1615025114 Zhao et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1

compared with other molecules, which reflects subtle changes inthe packing of the spine helix against the P domain β-strandsamong those molecules.Taken together, the Sf6 capsid protein exhibits remarkable

global conformational polymorphism. In particular, the drasticbending of the plank module matches the need to fit with thequasi-equivalent local environments in the icosahedral lattice. InHK97, the spine helix adopts bent conformations in the 3.65-Åresolution structure of prohead II (19, 20) but is straight in headII (11, 12), and the P domain in the HK97 prohead II showstwisting compared with that in head II. These conformationalchanges were proposed to be responsible for procapsid-to-capsidmaturation. Our results show that the P-domain β-sheet twistingand spine helix bending occur among the seven molecules inthe context of the authentic phage mature capsid, which wasnot seen in the previous HK97 head II structure at 3.44 Å res-olution (12). Additionally, the HK97 prohead II structure alsoshowed different conformations for the spine helix at differ-ent quasi-equivalent positions (20). Thus, each of the sevenmolecules undergoes a distinct conformational transition duringmaturation.An interesting feature of the observed conformational poly-

morphism in the plank module, including the E-domain bending,the P-domain β-sheet twisting, and the spine helix bending, isthat it cannot be described as a rigid-body movement, but insteadis propagated gradually over the length. We propose that theobserved plasticity of the plank modules including the spine helicesreflects conformational changes that are needed to accommodatethe densely packed DNA during the DNA-packaging process.Thus, the observed conformation likely represents a tensed stateunder the putative high internal pressure in a DNA-filled, au-thentic capsid. In contrast, the straight conformation of the spinehelix observed in the HK97 head II structure may represent a re-laxed state, consistent with the fact that the particles were assem-bled in vitro and did not contain phage DNA and hence nointernal pressure.

The Hexon Is Asymmetric and Is Curled. Molecules A to F form thehexon, one of the two types of capsomers in the Sf6 capsid. Thehexon is formed via A-domain contacts at the center and “head-to-tail” interactions between adjacent E loops and P domains atthe periphery (Fig. 1B and Fig. 2). The E loop from molecule Bforms extensive interactions with the P domain of molecule A. Inparticular, R60 of molecule B reaches into an indentation in theP domain of molecule A, where it makes a salt bridge withE123 in the spine helix (Fig. 2B). Residues 59–61 in the E loop ofmolecule B stack on the top of residues 86–88 in β3 of theP domain from molecule A, resulting in extension of theP-domain β-sheet by an additional β-strand. A mainchain:main-chain H-bond is formed between residue N72 in the E loop of

molecule B and A90 in the P domain of molecule A, forming aβ-sheet–like interaction. The I domain is engaged in numerousintra- and intermolecular interactions. Residues 321–327 act likea finger that pushes down the E loop in molecule B onto the Pdomain of molecule A. Loop 238–245 interacts with theN-terminal residue of the N arm from an adjacent capsomer.Loop 277–280 reaches the molecule A spine helix, whereas loop281–290 interacts with the I domain from molecule A. Such amode of intermolecular interactions repeats in each pair of ad-jacent molecules in the hexon.The hexon considerably deviates from a proper sixfold sym-

metry and, indeed, displays an approximate twofold symmetry(SI Appendix, Fig. S6 A and B). It was observed previously inHK97 and other phages that hexons in procapsid are twofoldskewed (15, 19, 21, 22), and it was proposed that they aretransformed into sixfold symmetric hexons in the virion duringthe maturation process (11, 19). The present high-resolutionstructure of Sf6 shows that maturation of the procapsid formedby highly skewed hexons does not result in sixfold symmetrichexons, but, instead, a considerable extent of skewing remains inthe hexons in mature capsid, which is manifested not only in themore flexible P domains and E loops but also in the A domainsand I domains.In addition to being asymmetric, the hexon has a curled shape

(SI Appendix, Fig. S6 C and D). In the context of the capsid, sucha curled shape results in a uniformly curved, smooth, innersurface for the five hexons surrounding a penton, which matcheswell with the curvature of the outermost layer of the packagedDNA that presumably pushes outward against the capsid andmay reflect the need to accommodate the packaged DNA andevenly propagate the high internal pressure due to the denselypacked DNA. We inspected the capsid protein residues that areclose to the outermost layer of the DNA. β22 and flanking loopsare closest to DNA and thus most likely make contact with DNA(SI Appendix, Fig. S6E). Interestingly, β22 is not present inHK97. Residues that are close to the DNA include several basicresidues—K29, R193, R209, K369, and R386—which may neu-tralize the negative charges of the DNA. Nevertheless, severalacidic residues—E106, E107, D189, D374, and E381—are alsoproximal to the DNA. The distribution of positively and negativelycharged residues at the capsid inner surface might impact orguide the organization of the DNA. However, a detailed un-derstanding of this requires the accurate structure of the DNA,which is absent in the current cryoEM map as it is icosahedrallyaveraged.

The T-Rich I Domain. Each phage particle is studded with 415copies of the I domain exposed on the capsid outer surface. The Idomain consists of a six-stranded β-barrel decorated with shortand long loops (Fig. 2B). The NMR structure of the phage P22I domain combined with lower-resolution cryoEM structuresshowed a fold similar to those of the folds in the translationinitiation factor eIF2gamma-family proteins (23). Consistentwith this, the six-stranded β-barrel of the Sf6 I domain resemblesthe folds of the translation initiation factor eIF2gamma-familyproteins, the P22 I domain, the Pseudomonas phage phi297tailspike gp61 distal domain, and the phage T5 distal tail proteinpb9 (SI Appendix, Fig. S6F). Interestingly, 18% of residues in theI domain are threonine, and those threonine residues frequentlyappear in TUT and TUTUT motifs, where U is a hydrophobicresidue, on the surface of the protein (Fig. 2B and SI Appendix,Fig. S3). The hoc protein decorating the outer surface of phageT4 capsid is also rich in threonine residues. The T4 hoc proteinhas been implicated in interaction with metazoa, probably viainteraction with mucin lining the mucus of the human intestine,thus reducing the attack of intestinal cells by host bacteriaof T4 and forming a line of non–host-derived innate immunity(24, 25). The Sf6 I domain may play a similar role by interactingwith the mucin in the human intestinal epithelium, thus in-creasing the abundance of the phage in the mucus and reduc-ing the infection by S. flexneri. Alternatively, the I domain may

A

R60, Mol B E123, Mol A

3, Mol A

N72, Mol B A90, Mol A

Loop 277-280

Loop 321-327

Loop 238-245

Loop 281-290

B

Fig. 2. Structure of the hexon. (A) Overview of the hexon shown as mo-lecular surface. The color scheme is the same as in Fig. 1B except that the Idomains are colored dark gray and light gray. In particular, molecule B isshown as a ribbon diagram in light blue, and its I domain is in brown. (B) Aview tilted by ∼45° about the horizontal axis from A. The I domain ofmolecule B is shown as a ribbon diagram in brown, and side chains ofthreonine residues are shown as stick models in red.

Zhao et al. PNAS | April 4, 2017 | vol. 114 | no. 14 | 3603

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

interact with surface features of Shigella flexneri (26), plausiblywith low affinity, thus promoting high-affinity adsorption of thetail to the bacterial host and increasing the rate of successfulinfection. Interestingly, S. flexneri can target key steps in mucinbiosynthesis and function including gene transcription, proteinglycosylation, and secretion as strategies to alter the mucusbarrier to facilitate infection (27). A common feature of mucinand the surface of S. flexneri is carbohydrates, as mucin is highlyglycosylated and S. flexneri is covered by lipopolysaccharides. Itwill be interesting to see if and how these structural features inthe I domain relate to the interplay among the phage, its bac-terial host, and the metazoan. In phage P22, the I domain wasshown to play roles in capsid protein folding, capsid assembly,and incorporation of the portal protein (23). It is not known ifthe Sf6 I domain also plays these functional roles.

Molecular Determinants That Mediate Differential Formations of theHexon and the Penton. Pentons introduce the curvature into theprotein shell and are essential for icosahedral capsid assembly. Itwas not previously understood how the two types of capsomers,i.e., pentons and hexons, are usually formed by the same capsidprotein in tailed DNA phages. The present phage Sf6 structurereveals the structural basis in atomic detail.In Sf6, the penton is formed with an overall similar pattern of

intermolecular interactions as the hexon; that is, five copies of Adomains juxtapose to form the central core, surrounded byflexible plank modules that are more bent than those in thehexon (Fig. 3 and SI Appendix, Fig. S7). However, the relativeorientation between adjacent molecules in the penton changesdrastically by ∼39° compared with that in the hexon, therebygenerating the curvature (SI Appendix, Fig. S7 A–D). Thischange is mediated mainly by A domains, and the details of in-termolecular interactions differ significantly between pentonsand hexons. Residues 182–206 adopt drastically different con-formations between the hexon and penton. Within this region,residues 189–198 form an α-helix near the center of the hexon,whereas they refold to form a β-structure in the penton, adding astrand to the β-sheet in the A domain (Fig. 3 and Movie S2).Accompanying this α–helix-to-β–strand transition, residue W196moves by as much as 16 Å. In the hexon, W196 stacks with W146,whereas in the penton W196 moves away from W146 and insertsinto a hydrophobic pocket in the adjacent molecule formed byα3, β7, and the newly formed β-strand (SI Appendix, Fig. S7 Eand F). In the hexon, the W146 side chain inserts betweenα3 from one molecule and α4 from the adjacent molecule,whereas in the penton it flips over, dragging helix α4 closer toα3 and causing a rotation of helix α4. A third tryptophan residue,W173, is located at the outermost periphery of the A domain(distal to the capsomer center) and is inserted into a hydrophobic

crevice formed by α3 and loop 281–290 from the adjacent mol-ecule, which involves similar intermolecular interactions in thehexon and the penton, thereby serving as a hinge to hold togetherthe outermost periphery of the two adjacent A domains. Con-sequently, the A domain in the penton rolls along its adjacentcounterpart, leading to significant bending between the twoneighboring molecules, compared with the hexon, and generatingthe curvature. Due to the bending, the five copies of I domains inthe penton, which are located at higher radii, have few inter–I-domain interactions. There are ∼20% more intermolecularH-bonds between a pair of neighboring molecules in the pentonthan in the hexon (∼50% increase for intermolecular H-bondsinvolving A domains), suggesting that the penton is significantlymore stable than the hexon in the context of the capsid, whichis consistent with its role in generating the curvature for theicosahedral lattice.Such a α–helix-to-β–strand transition was not observed in

HK97 (12). Instead, the portion in HK97 corresponding to theSf6 helix α5 (residues 189–198) is a loop that exhibits differentialconformations between the hexon and penton environments,likely serving as a similar structural determinant that mediatesdifferential formation of hexons and pentons. Additionally, inHK97, A domains from adjacent molecules show multiplecharge:charge interactions involving basic and acidic residues(12). Such charge:charge interaction is absent in Sf6.

Adjacent Capsomers Are Interlocked via Mutually PenetratingStructural Elements, Resembling Mortise-and-Tenon Joinery. Adja-cent capsomers are interlocked via mutually penetrating struc-tural elements from plank modules and N arms, reminiscent ofmortise-and-tenon joinery (Fig. 4 and SI Appendix, Fig. S8). TheP-domain β-sheet and the E loop from an adjacent molecule inthe same capsomer form a pocket into which helix α1 and theP loop (residues 393–396) in the juxtaposing molecule fromthe other capsomer insert, forming numerous intercapsomerH-bonds and representing a three-molecule synapse. The overallmode of these interactions resembles that in HK97 (12), but thedetails differ significantly (Fig. 4 and SI Appendix, Fig. S8A). Themortise-and-tenon–like intercapsomer interactions exhibit pre-cise structural complementarity at the interface, which not onlyconnects adjacent capsomers strongly, but also can preventcapsomers from disengaging radially.

The N Arms Form a Network Throughout the Capsid. The N arm ishighly extended, forming abundant interactions with neighboringmolecules along a path length of 167 Å (Fig. 5 and SI Appendix,Fig. S9), and such interactions frequently involve a single residueor a short polypeptide fragment being engaged with multiplemolecules within a single capsomer or from different capsomers.The very N terminus of the N arm is exposed on the outer sur-face of the capsid. Using the N arm from molecule A as an ex-ample, residues N3 and N4 juxtapose with I75 and S76 in the Eloop of molecule F from a different capsomer, forming β–β in-teractions by making H-bonds between mainchain oxygen andnitrogen atoms (Fig. 4C and Fig. 5). Residue N3 side chainmakes H-bonds with E92 in the P domain of molecule E and withL74 in the E loop of molecule F, both from the adjacent cap-somer, thus contributing to the stability of the adjacent cap-somer. Residue S7 side chain is H-bonded to the side chain ofD243 in the I domain in molecule F from the adjacent capsomer.The N arm then reaches the quasi- or icosahedral twofold axis,where it forms β–β interactions with its equivalent in molecule Efrom a different capsomer by juxtaposed residues S10 and Q11.This fragment also forms van der Waals interactions with resi-dues 43–45 of the N arm in molecule B from the same capsomer(Fig. 5 and Fig. 4C). The N arm then extends into the capsidinterior and runs in an elongated crevice underneath moleculesB and A (SI Appendix, Fig. S9). Residue F21 side chain insertsinto a pocket between the A and I domains of molecule B. TheC-terminal end of the N arm finally arrives at a site next to an-other quasi- or icosahedral twofold axis, which is related to the

A B

W196W146

W146

W173

W196

W173

Helix 189-198

Fig. 3. Structural basis of differential formations of the penton and thehexon. (A) The penton. The five molecules are rainbow-colored from blue tored. Residues 182–206 are in brown. (B) The hexon. Residues 182–206 are ingreen. In A and B, side chains of residues W146, W173, and W196 (magenta)as well as surrounding hydrophobic residues (gray) are shown as stickmodels. (Inset) Transition from α-helix (dark green) to β-strand (orange) forresidues 189–198. The molecule from the penton and the hexon is in red andlight blue, respectively.

3604 | www.pnas.org/cgi/doi/10.1073/pnas.1615025114 Zhao et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1

previous one by a quasi- or icosahedral threefold symmetry,where it interacts with helix α1 in the P domain of molecule Gfrom the adjacent penton and the N-terminal fragment of the Narm of molecule F from the same capsomer (Fig. 5 and SI Ap-pendix, Fig. S9). Therefore, each N arm interacts with as many assix molecules, that is, molecules A, B, and F from the samecapsomer, molecules E and F from the adjacent capsomer, andmolecule G from the adjacent pentameric capsomer. At eachquasi- or icosahedral twofold axis, four N arms join to form amolecular junction. In the context of the capsid, the N armsinterweave into a network that tangentially holds together cap-somers (Fig. 5B).It is interesting to note that the β–β interactions between two

juxtaposed N arms (residues S10 and Q11) at the quasi- andicosahedral twofold axes as described above for Sf6 are missingin the HK97 structure. In HK97, intermolecular interactions attwofold axes involve four glutamine residues that are arrangedloosely in a head-to-head manner at the hexon:hexon interfacebut in a more compact, interlocked, zipper-like manner at thepenton:hexon interface (12). It is likely that the β–β interactions,that is, mainchain:mainchain H-bonds, contribute to strongerintermolecular interactions; thus, higher stability of theSf6 capsid and the absence of such interaction in HK97 maycorrelate with the requirement for the unique covalent cross-linking between residues K169 in the E loop and N356 in theP domain. Given the fact that most tailed phages do not have thecovalent cross-linking as in HK97, it is likely that such β–β in-teraction may be present in other tailed phages.In Sf6, each N arm from hexons is gripped deeply in a narrow

crevice between a pair of adjacent molecules (SI Appendix, Fig.S9). The N arms from pentons are more buried due to thebending intermolecular interface (SI Appendix, Fig. S9). Thedepth and length of those crevices and the nature of the exten-sive molecular interactions formed by N arms suggest that it isunlikely for N arms to move into those narrow and deep crevicesafter those crevices are formed. Instead, N arms may have oc-cupied the observed positions before formation of the hexonsand pentons and before formation of the mature capsid. More-over, the present structure shows precise structural comple-mentarity at the intercapsomer interfaces around the quasi- or

icosahedral twofold axes, leaving no space for the veryN-terminal region to be threaded from the capsid interior to theexterior after such intercapsomer interfaces are formed. This iseven more unlikely in phage T7, the capsid protein of which hasa longer N-terminal fragment that is also exposed on the maturecapsid exterior (15). Previous results showed that, in procapsid,N arms are situated at the inner surface of the protein shell,putatively kept in a metastable state by interaction with thescaffold proteins (15, 19, 21, 22, 28). It is likely that, duringprocapsid-to-capsid maturation, N arms depart from the originalstate as seen in the procapsid and begin to bind to interactingpartners and extend to the outside of the protein shell, whichpromotes conformational changes and capsid protein rear-rangements that lead to formation of mature forms of hexonsand pentons and ultimately the mature capsid. Our results sug-gest that N arms may play an active role in regulating capsidassembly and maturation.

Implications for Capsid Assembly. The present structure reveals, atatomic detail, assembly of a headful DNA-packaging phage in itsauthentic, DNA-filled state. Capsomers are assembled with rel-atively invariant A domains forming the central cores, which aresurrounded by highly flexible peripheral domains formingintercapsomer interfaces (SI Appendix, Fig. S10). Differentialformations of hexons and pentons are mediated by a dramaticstructural rearrangement in the A domain, leading to curvatureat the fivefold vertices that is required for formation of theicosahedral lattice. Intercapsomer interfaces are characteristic ofprecise, mortise-and-tenon–like interactions. The highly ex-tended N arms interweave into a network over the capsid andtangentially hold together capsomers and render the capsidultrastable. The N arms likely play an active role in regulatingcapsid assembly and maturation. Additional domains, in this casethe threonine-rich I domains, are studded on the capsid outersurface and take part in interactions with hosts or metazoa. Suchan assembly scheme may be common among tailed DNA phagesand herpesviruses.The seven chemically identical capsid protein subunits in the

icosahedral asymmetric unit are all structurally distinct, indicat-ing remarkable global conformational polymorphism. The plankmodules undergo bending to different extents among the sevenmolecules, dependent on their local environments in the icosa-hedral lattice. The hexon is asymmetric and curled, resulting ina curved, smooth (instead of faceted) inner surface for the five

capsomer 2 capsomer 1

A

B

C

D394

L96

I100

K71

K399

I66

L405 V83

N4

N3 I75

S76

capsomer 2 capsomer 1

Fig. 4. The intercapsomer interface. (A) An intercapsomer interface viewedfrom the exterior. Molecules in the capsomer on the left are colored as inFig. 1B. N arms are shown as stick models. The N arms of molecule B and Aare in magenta and dark blue, respectively. Molecules in the capsomer onthe right are colored in yellow except that N arms are in gray. Only ap-proximately half of each capsomer is shown. The approximate locations oficosahedral threefold axes are indicated with black triangles. (B) Themortise-and-tenon intercapsomer interaction (molecular surface). The viewis approximately tangential to the capsid surface. The color scheme is thesame as in A. (C) Closeup of the intercapsomer interaction. IntermolecularH-bonds are shown as dashed lines in dark green.

A

F21

S10 Q11

S43

1

F21

B

Fig. 5. The N arm network. (A) The N arm junction at an intercapsomerinterface. The color scheme is the same as in Fig. 4A. N arms are shown asstick models with side chains. The N arm of molecule A (blue) is in light cyan.The N arm of molecule B (light blue) is in magenta. The N arms of moleculesfrom an adjacent capsomer (ribbon diagram in yellow) are gray. The loca-tions of the icosahedral threefold axes are indicated by black triangles. Thelocal twofold axis is indicated by a black oval near the center of the picture.H-bonds are shown as gray dashed lines. (B) N arms (sphere models in purple)form a network throughout the capsid. Other portions of capsid proteins areshown as ribbon diagrams in gray. The seven molecules in one asymmetricunit are rainbow-colored. Three N-arm junctions are indicated with greenellipsoids, one of which corresponds with the close-up in A.

Zhao et al. PNAS | April 4, 2017 | vol. 114 | no. 14 | 3605

BIOCH

EMISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

hexons surrounding a penton. Curling of the hexons and bendingof the plank modules make the capsid deviate from the idealicosahedron. Indeed, the Sf6 capsid is estimated to have a spherefactor of ∼0.5; that is, the capsid is approximately halfway betweenan ideal icosahedron and a sphere. On the one hand, these fea-tures match the structural requirements for formation of the T = 7icosahedral protein shell. On the other hand, they may reflect theneed for uniformly propagating the putative high internal pressureonto the protein shell wall. The observed flexibility of the capsidproteins may reflect conformational changes that are needed toaccommodate the densely packed DNA during the DNA-packagingprocess. Thus, the present structure likely represents a tensed stateunder the putative high internal pressure in an authentic, DNA-filled capsid, mimicking an inflated balloon.The capsid is assembled via noncovalent interactions, but

displays ultrastability as required to stand with the high internalpressure resulting from DNA packaging. Such stability is madepossible by several features of intermolecular interactions in theSf6 capsid. The extended N arms provide a long-distance rein-forcing mechanism for the capsid, and portions of the N armsoften lie at molecular interfaces and are frequently engaged inmultivalent interactions with multiple molecules. A number of β–βintermolecular interactions and mainchain:mainchain H-bondsare observed. The β–β interactions are reminiscent of those infibroin in silkworm silk (29) and spidroin in spider silk (30), whichpossess outstanding physical stiffness/toughness and are arguablyamong the strongest, yet flexible, proteinaceous materials innature. The β–β interactions are also dominant in β-amyloid (31).We propose that the abundant β–β interactions observed in theSf6 structure confer significant physical strength yet flexibilityas required for dynamic capsid assembly and DNA packaging.This may explain why the procapsid-to-capsid maturation processis ubiquitous among tailed DNA phages. A metastable procapsidforms first, aided by scaffold proteins, which then transforminto the highly stable mature capsid in an irreversible process,triggered by DNA packaging and mediated by structural

elements such as the N arms. The advantage of such a two-stepassembly mechanism over direct formation of the ultrastablecapsid is to prevent the formation of aberrant products caused byan avalanche of formation of strong interactions as observed inthe mature capsid.Overall, the present structure delineates how a DNA-inflated,

thin, rigid, protein shell is constructed via noncovalent interac-tions. The structure can have implications in fabricationof polypeptide-based, strong but flexible, self-assembled nano-containers.

Materials and MethodsThe Sf6 phage was grown and purified as described in SI Appendix, SI Text.The purified phage was frozen-hydrated using a standard procedure. ThecryoEM images were recorded on a Titan Krios with a K2 Summit directelectron detector operating on a superresolution counting mode. A total of2,009 movies were recorded, from which 69,094 particles were boxed. Thestructure was determined with single-particle cryoEM method using the jsprsoftware package (32). The atomic model was built automatically withPhenix, aided by manual rebuilding with Coot. The model was refined in thereal space with Coot and then in the reciprocal space with Phenix usingstructure factors calculated from the region covering the asymmetric unitcarved from the cryoEM map. The refined model is of good stereochemistry(SI Appendix, Table S1). The final model contains residues 2–423 for all of theseven protein molecules in the icosahedral asymmetric unit. The only missingresidue is the N-terminal Met. There is a density at the center of the hexon,surrounded by the six-histidine residue H186, which was modeled as achloride ion due to its abundance in the sample buffer.

ACKNOWLEDGMENTS. We thank the Purdue Cryo-EM Facility for providingthe data collection resource, the Purdue Rosen Center for Advanced Com-puting and the University of Kansas Research Computing for the computationalresource, and Mr. Steve Wilson and Dr. Lev Gorenstein for assistance in datatransferring using Globus. This work was supported by Grant R01GM090010, theKansas Idea Network of Biomedical Research Excellence Grant P20 GM103418 (toL.T.), and Grant R01AI111095 (to W.J.) from the National Institute of GeneralMedical Sciences of the National Institutes of Health.

1. Suttle CA (2007) Marine viruses: Major players in the global ecosystem. Nat RevMicrobiol 5(10):801–812.

2. Hendrix RW (2003) Bacteriophage genomics. Curr Opin Microbiol 6(5):506–511.3. Sun S, et al. (2008) The structure of the phage T4 DNA packaging motor suggests a

mechanism dependent on electrostatic forces. Cell 135(7):1251–1262.4. Zhao H, Christensen TE, Kamau YN, Tang L (2013) Structures of the phage Sf6 large

terminase provide new insights into DNA translocation and cleavage. Proc Natl AcadSci USA 110(20):8075–8080.

5. Fokine A, Rossmann MG (2014) Molecular architecture of tailed double-stranded DNAphages. Bacteriophage 4(1):e28281.

6. Smith DE, et al. (2001) The bacteriophage straight phi29 portal motor can packageDNA against a large internal force. Nature 413(6857):748–752.

7. Smith DE (2011) Single-molecule studies of viral DNA packaging. Curr Opin Virol 1(2):134–141.

8. Grayson P, et al. (2006) The effect of genome length on ejection forces in bacterio-phage lambda. Virology 348(2):430–436.

9. Gelbart WM, Knobler CM (2009) Virology. Pressurized viruses. Science 323(5922):1682–1683.

10. Suhanovsky MM, Teschke CM (2015) Nature’s favorite building block: Decipheringfolding and capsid assembly of proteins with the HK97-fold. Virology 479-480:487–497.

11. Wikoff WR, et al. (2000) Topologically linked protein rings in the bacteriophageHK97 capsid. Science 289(5487):2129–2133.

12. Helgstrand C, et al. (2003) The refined structure of a protein catenane: TheHK97 bacteriophage capsid at 3.44 A resolution. J Mol Biol 334(5):885–899.

13. Zhou ZH, et al. (2014) Four levels of hierarchical organization, including noncovalentchainmail, brace the mature tumor herpesvirus capsid against pressurization.Structure 22(10):1385–1398.

14. Baker ML, Jiang W, Rixon FJ, Chiu W (2005) Common ancestry of herpesviruses andtailed DNA bacteriophages. J Virol 79(23):14967–14970.

15. Guo F, et al. (2014) Capsid expansion mechanism of bacteriophage T7 revealed bymultistate atomic models derived from cryo-EM reconstructions. Proc Natl Acad SciUSA 111(43):E4606–E4614.

16. Gemski P, Jr, Koeltzow DE, Formal SB (1975) Phage conversion of Shigella flexnerigroup antigens. Infect Immun 11(4):685–691.

17. Casjens S, et al. (2004) The chromosome of Shigella flexneri bacteriophage Sf6:Complete nucleotide sequence, genetic mosaicism, and DNA packaging. J Mol Biol339(2):379–394.

18. 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(2):207–215.

19. Gertsman I, et al. (2009) An unexpected twist in viral capsid maturation. Nature458(7238):646–650.

20. Gertsman I, Komives EA, Johnson JE (2010) HK97 maturation studied by crystallog-raphy and H/2H exchange reveals the structural basis for exothermic particle transi-tions. J Mol Biol 397(2):560–574.

21. Chen DH, et al. (2011) Structural basis for scaffolding-mediated assembly and matu-ration of a dsDNA virus. Proc Natl Acad Sci USA 108(4):1355–1360.

22. Parent KN, et al. (2010) P22 coat protein structures reveal a novel mechanism forcapsid maturation: Stability without auxiliary proteins or chemical crosslinks.Structure 18(3):390–401.

23. Rizzo AA, et al. (2014) Multiple functional roles of the accessory I-domain of bacte-riophage P22 coat protein revealed by NMR structure and CryoEM modeling.Structure 22(6):830–841.

24. Barr JJ, et al. (2015) Subdiffusive motion of bacteriophage in mucosal surfaces in-creases the frequency of bacterial encounters. Proc Natl Acad Sci USA 112(44):13675–13680.

25. Barr JJ, et al. (2013) Bacteriophage adhering to mucus provide a non-host-derivedimmunity. Proc Natl Acad Sci USA 110(26):10771–10776.

26. Simpson DJ, Sacher JC, Szymanski CM (2015) Exploring the interactions betweenbacteriophage-encoded glycan binding proteins and carbohydrates. Curr Opin StructBiol 34:69–77.

27. Sperandio B, et al. (2013) Virulent Shigella flexneri affects secretion, expression, andglycosylation of gel-forming mucins in mucus-producing cells. Infect Immun 81(10):3632–3643.

28. Veesler D, et al. (2014) Architecture of a dsDNA viral capsid in complex with itsmaturation protease. Structure 22(2):230–237.

29. Leal-Egaña A, Scheibel T (2010) Silk-based materials for biomedical applications.Biotechnol Appl Biochem 55(3):155–167.

30. Rising A, Johansson J (2015) Toward spinning artificial spider silk. Nat Chem Biol 11(5):309–315.

31. Sawaya MR, et al. (2007) Atomic structures of amyloid cross-beta spines reveal variedsteric zippers. Nature 447(7143):453–457.

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

3606 | www.pnas.org/cgi/doi/10.1073/pnas.1615025114 Zhao et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1