X-ray Crystal Structure of the Rotavirus Inner Capsid Particle ...crystal.harvard.edu/wp-content/uploads/2018/12/mcclain.pdfof polypeptide chain (residues 81–100) leading into them

doi:10.1016/j.jmb.2010.01.055 J. Mol. Biol. (2010) 397, 587–599

Available online at www.sciencedirect.com

X-ray Crystal Structure of the Rotavirus Inner CapsidParticle at 3.8 Å Resolution

Brian McClain1, Ethan Settembre1, Brenda R. S. Temple1,A. Richard Bellamy2 and Stephen C. Harrison1,3⁎

1Laboratory of MolecularMedicine, Children's HospitalBoston, 320 Longwood Avenue,Boston, MA 02115, USA2School of Biological Sciences,University of Auckland,Auckland 1142, New Zealand3Howard Hughes MedicalInstitute, 320 LongwoodAvenue, Boston, MA 02115,USA

Received 8 December 2009;received in revised form19 January 2010;accepted 22 January 2010Available online1 February 2010

0022-2836/$ - see front matter © 2010 E

*Corresponding author. E-mail addPresent addresses: B. McClain, Ve

Diagnostics, Cambridge, MA, USA;Abbreviations used: DLP, double-

particle; cryo-EM, cryo-electron mic

The rotavirus inner capsid particle, known as the “double-layered particle”(DLP), is the “payload” delivered into a cell in the process of viral infection. Itsinner and outer protein layers, composed of viral protein (VP) 2 and VP6,respectively, package the 11 segments of the double-stranded RNA (dsRNA)of the viral genome, as well as about the same number of polymerasemolecules (VP1) and capping-enzyme molecules (VP3). We have determinedthe crystal structure of the bovine rotavirus DLP. There is one full particle(outer diameter ∼700 Å) in the asymmetric unit of the P212121 unit cell ofdimensions a=740 Å, b=1198 Å, and c=1345 Å. A three-dimensionalreconstruction from electron cryomicroscopy was used as a molecularreplacement model for initial phase determination to about 18.5 Å resolution,and the 60-fold redundancy of icosahedral particle symmetry allowed phasesto be extended stepwise to the limiting resolution of the data (3.8 Å). Thestructure of a VP6 trimer (determined previously by others) fits the outer layerdensity with very little adjustment. The T=13 triangulation number of thatlayer implies that there are four and one-third VP6 trimers per icosahedralasymmetric unit. The inner layer has 120 copies of VP2 and thus 2 copies pericosahedral asymmetric unit, designated VP2A and VP2B. Residues 101–880fold into a relatively thin principal domain, comma-like in outline, shaped suchthat only rather modest distortions (concentrated at two “subdomain”boundaries) allow VP2A and VP2B to form a uniform layer with essentiallyno gaps at the subunit boundaries, except for a modest pore along the 5-foldaxis. The VP2 principal domain resembles those of the corresponding shellsand homologous proteins in other dsRNA viruses: λ1 in orthoreoviruses andVP3 in orbiviruses. Residues 1–80 of VP2A and VP2B fold together with fourother such pairs into a “5-fold hub” that projects into theDLP interior along the5-fold axis; residues 81–100 link the 10 polypeptide chains emerging from a 5-fold hub to the N-termini of their corresponding principal domains, clusteredinto a decameric assembly unit. The 5-fold hub appears to have several distinctfunctions. One function is to recruit a copy of VP1 (or of a VP1–VP3 complex),potentially along with a segment of plus-strand RNA, as a decamer of VP2assembles. The second function is to serve as a shaft around which can coil asegment of dsRNA. The third function is to guide nascent mRNA, synthesizedin theDLP interior byVP1 and 5′-capped by the action of VP3, out through a 5-fold exit channel. We propose a model for rotavirus particle assembly, basedon known requirements for virion formation, together with the structure of theDLP and that of VP1, determined earlier.

© 2010 Elsevier Ltd. All rights reserved.

Keywords: double-layered particle; double-stranded RNA; virus assembly;icosahedral symmetry
Edited by I. Wilson
lsevier Ltd. All rights reserved.

ress: [email protected] Pharmaceuticals, Cambridge, MA, USA; E. Settembre, Novartis Vaccines andB. R. S. Temple, University of North Carolina, Chapel Hill, NC, USA.layered particle; VP, viral protein; dsRNA, double-stranded RNA; ICP, inner capsidroscopy; SLP, single-layered particle.

http://dx.doi.org/10.1016/j.jmb.2010.01.055mailto:[email protected]

588 Structure of the Rotavirus Inner Capsid Particle

Introduction

Viruses with double-stranded RNA (dsRNA)genomes are found among infectious agents ofanimals, plants, yeasts, and bacteria.1 Except forsome dsRNA viruses of bacteria and fungi, virionsare generally multilayered, icosahedrally symmet-ric, nonenveloped particles. Despite the hugediversity of their hosts, these viruses share a numberof genomic and structural characteristics, includinga form of the virus particle (in most cases, a subviralparticle stripped of one or more outer shells) thatbecomes a transcriptionally active, mRNA-produc-ing “molecular machine” when transferred to thecytoplasm of a host cell.The transcriptionally active inner capsid particle

(ICP) of these viruses has various designations,depending on the history of nomenclature for theparticular group of dsRNA viruses in question. TheICPs of orthoreoviruses and orbiviruses are knownas “core” particles; those of rotaviruses (Fig. 1) areknown as “double-layered particles” (DLPs). Thecrystal structures of the reovirus and orbivirus coresand of the yeast L-A virus have been described,2,3 ashave cryo-electron microscopy (cryo-EM) recon-structions of several other dsRNA particles, atvarious resolutions.4–7 All but the birnaviruses8

appear to have in common a shell (composed of120 copies of a related but quite variable protein)immediately surrounding the tightly coiled dsRNA.Other components can be quite different. Forexample, orthoreovirus cores (such as the ICPs ofaquareoviruses and oryzaviruses) have projecting 5-fold “turrets,” which are capping enzymes thatmodify viral mRNA as it passes out of the particle.2

We report here the crystal structure of therotavirus DLP, determined with data that extend

Fig. 1. Cut-away view of the complete rotavirus virion(TLP), based on cryo-EM reconstructions and assignmentsof densities. The two outer layer proteins are shown inyellow (VP7) and red (VP5⁎/VP8⁎). The middle layerprotein (VP6) is shown in green; the inner layer VP2 isshown in blue. Coils of dsRNA (not shown here) fill theinterior, organized in part by internal structures of the ICPdescribed in this work.

to a minimum Bragg spacing of 3.8 Å. The DLP innershell protein is called viral protein (VP) 2. A layer ofVP6 trimers,9 organized in a T=13l icosahedrallattice, surrounds the 120-subunit VP2 layer. Withinthe VP2 shell are 11 genomic segments (varying inlength from ∼700 to ∼3100 nucleotide pairs) and 11or 12 copies each of the viral polymerase VP1 andthe viral capping enzyme VP3. These last twocomponents are not detected in the crystal structure.

Results

Structure determination

The specific procedures (experimental and com-putational) for crystallographic structure determi-nation are described in Materials and Methods.The crystals in space group P212121 with unit-celldimensions a=740 Å, b=1198 Å, and c=1345 Åcontain a full particle per asymmetric unit. The 60-fold noncrystallographic redundancy permitted arobust extension of low-resolution phases from an18.2-Å-resolution cryo-EM-based map to the full3.8-Å resolution of the data; it also allowed us toovercome substantial diffraction anisotropy. Wecould trace the VP2B polypeptide chain fromresidue 81 to the C-terminus at residue 880without much difficulty and, likewise, residues100–800 in VP2A; VP2A residues 81–100 were notas well defined, but we could follow the path ofthe main chain. We could also identify residues18–31 in an α-helix clustered near the 5-fold axiswith four symmetry mates, part of the inward-projecting feature we call the “5-fold hub.” Theknown structure of VP6 could be placed in themap with only local adjustments.

VP2

Most of the VP2 polypeptide chain folds into abroad plate-like structure with a somewhat bean-shaped or inverted-comma-shaped outline, 140 Ålong by 60 Å wide and varying in thickness from15 Å to 30 Å (Fig. 2a and b). Five monomers ofVP2 (conformer A: VP2A) form a star-like complexabout each 5-fold axis, and five additional VP2monomers (conformer B: VP2B) pack into the gapsbetween the points of the star. Twelve of thedecamers thus formed make up the shell of theICP (Fig. 2c). Of the 880 residues in the polypep-tide chain, only those from residues 81 to 880 canbe traced clearly. Residues 81–100 of VP2B form aloop and a helix that pack beneath a neighboringVP2A; the density extending inward from residue100 of VP2A indicates that a similar loop–helixelement is present, but not as firmly fixed by itscontacts with a neighboring subunit (Fig. 2d).Additional density features surrounding the 5-foldaxis at smaller radii show that the N-terminalsegments of VP2A and VP2B form a distinct hub-like assembly (see the text below).

Fig. 2. Structure of VP2. (a) Ribbon representation, with secondary structural elements labeled. Helices are shown indark blue; strands are shown in light blue. (b) Amino acid sequence (single-letter code) of UK Bovine VP2 (serogroup A)with secondary structural elements shown above the sequence. (c) The VP2 shell, viewed from outside the particle into the5-fold channel. The central decamer is highlighted, with VP2A shown in dark blue andwith VP2B shown in light blue. (d).Internal view of the VP2 shell, represented as a surface rendering with colors for VP2A and VP2B as in (c). Residues 81–100 (the last parts of the N-terminal arms) are represented, respectively, by white and gray “worms.”

589Structure of the Rotavirus Inner Capsid Particle

The shape of the VP2 subunit is such thatrelatively little distortion of one conformer is neededto create the other. Comparison of the A and Bconformers suggests the following description of the

internal flexibility in VP2. If one divides the subunitinto three regions, then there are modest hingemotions at the two interregional boundaries (Fig. 3).We designate these three regions as “apical,”

Fig. 3. Conformational flexibility of VP2 and VP2-like proteins: rotavirus VP2 (this work), BTV VP3,3 and reovirus λ1.2(a–c) In each panel, there are orthogonal views of the monomer closer to the fivefold (e.g., VP2A; in red and dark blue),superposed on the monomer farther from the fivefold (e.g., VP2B; in green and light blue). The proteins were aligned ontheir central domains. Rotavirus VP2 and BTV VP3 each has similar relative shifts of their dimer-forming domains (lowerpart of each figure; rmsd of 2.4 Å and 4.7 Å, respectively), but substantially different shifts of their apical domains (upperpart of each figure; 2.1 Å and 7.9 Å, respectively). λ1 has only two domains, with an rmsd of 5.4 Å in one when the other isaligned.


“central,” and “dimer-forming” (see Fig. 2c). If themiddle regions of the two conformers are aligned,the rmsd of the α carbon atoms in the apical regionsis 4.7 Å and that of the α carbons in the dimer-forming regions is 7.9 Å (Fig. 3a).The structure of VP2 resembles, as expected, those

of homologs in other dsRNA,viruses,2,3 but each ofthe known exemplars has special features, and it isnot evident from casual inspection how best to alignthem. Comparison of secondary structural elementsshows that the principal helices and sheets in thecentral and apical regions are conserved (Fig. 3).This conserved core is decorated with additionalelements, some of which are common to a subset ofthe proteins, and some of which are uniqueinsertions into the basic structure. Rotavirus VP2appears to have the smallest number of insertions.Despite a low sequence identity (b10%) among theseproteins, alignment of VP2 to the other structuresyields positional rmsd values of less than 5 Å for80% of each polypeptide chain. Descriptions of therelationships between the two conformers aresimilar for the three examples shown in Fig. 3,although in the case of reovirus λ1, there appear tobe only two independently superposable regions,with hinge discontinuity in the middle of thesubunit.The inward face of the VP2 shell is of roughly

uniform radius (240±5 Å), except for internalprojections near the 5-fold axes and the segmentsof polypeptide chain (residues 81–100) leading intothem. The charge distribution is unremarkable,with a slight excess of negatively charged sidechains. The capsid thus presents to the genomicRNA a boundary surface that is relatively smoothin both its steric features and its electrostaticfeatures (Fig. 2d).The convergence of five VP2A subunits around a

5-fold axis leaves only a relatively small pore, linedwith positive charge (Fig. 4). Flexion of the loopsthat impinge on this pore would be needed for RNA

transcripts to pass through. Some differences inthese loops in both VP2A and VP2B suggest thatthey can indeed adopt alternative conformations.

VP6

The VP6 subunit is a jelly-roll β-barrel withradially directed strands and with N-terminal andC-terminal extensions that together form a largely α-helical pedestal (Fig. 5a).9 Both domains participatein very extensive trimer contacts. There are fivedistinct positions for the trimer on the outer surfaceof the DLP. They are designated as T, T′, P, P′, and Din Fig. 5b. Only the first of these (T) lies on a particle3-fold axis. The conformation of VP6 in each of thefive locations is essentially the same as in crystals ofthe free trimer, with the exception of some variationsin the segment containing residues 64–71, which is aprincipal contact with VP2.The intertrimer contacts have a local 2-fold

character, as expected from their quasi-equivalentpacking. Because the trimers project radially, thecontacts involve residues in the pedestal domain,not in the apical β-barrel. There are seven distincttrimer–trimer contacts, one of which is across a stricticosahedral dyad (Fig. 5, D–D). They are all ofmodest extent, resembling more the interfaceswithin protein crystals than those within tightlyknit assemblies. At five of the seven, the loopscontaining residues 142–146 approach each otheracross the local twofold, and a small network ofpolar interactions links the apposed trimers (Fig. 6).The contributing residues Gln142, Asn143, andArg145 are conserved in serogroup A and serogroupC rotavirus strains (e.g., see Heiman et al.10), as isAsp380, which salt bridges to Arg145. Also con-served are the partners in an additional salt bridge,somewhat displaced from the local twofold but atabout the same radial position, between Arg106 andGlu353. Two of the trimer contacts have alternativesalt-bridge patterns. At the D–P′ interface, the

Fig. 4. The 5-fold channel formed by VP2A monomers(backbone ribbons), with key conserved positivelycharged residues (Lys526, Arg527, and Arg531) shownas ball-and-stick representations. The VP2 loops (residues354–363) that form an internal gate are highlighted in red.


apposed trimers are displaced laterally by about10 Å, with respect to their relative positions at thefive nearly invariant interfaces, and Arg106 (nowclose to the local twofold) interacts with Glu379,Asp380, and Gln383 (Fig. 6). The P–P interface isdistinctive because thickening of the VP2 layer nearthe fivefold displaces the P trimers radially outwardand tips the trimer 3-fold axes away from the

Fig. 5. Structure and packing of VP6. (a) Ribbon representatto their positions with respect to the icosahedral symmetry axblue (D; closest to the twofold), light blue (P; closest to the fiveshown in red and green coils.

fivefold (Fig. 6). At these contacts, Arg106 engagesin a salt bridge with Glu332 (and perhaps also withGlu379), while the 142–146 loops are slightlydisplaced from each other. Thus, the adaptabilityof salt bridges between relatively long and con-served side chains (Arg and Glu) permits quasi-equivalent adjustments at five of the intertrimerinterfaces, while ion-pair switching among con-served alternatives occurs at the remaining twointertrimer interfaces. Comparison of the VP6 trimerstructure with a moderate-resolution cryo-EM re-construction has led to similar, although lessdetailed, conclusions.9

VP2–VP6 contacts

The principal interaction occurs between theinward-projecting β2–α3 loop (residues 64–72) ofVP6 and the outer surface of VP2. The β2–α3 loop,an extended segment of polypeptide chain that runsalongside two α-helices, manages to adjust to the 13distinct VP2 locations that it faces, so that the sidechains of Thr69 and Leu71 make some sort of vander Waals contact in each. The most intimate contactoccurs beneath VP6-T: the side chains of Thr69 andLeu71 fit into a pocket formed by Met228, Met839,Met841, Phe244, and Ala220 of VP2B (Fig. 7). Thecorresponding pocket on the surface of VP2Areceives the side chains of Leu70 and Leu71 fromone of the three subunits of VP6-D. Beneath VP6-P′,one of the β2–α3 loops has no contact with the VP2layer, but the other two subunits and the interveningsurface have a very extensive interaction with VP2B,including a network of hydrogen bonds among side-

ion.9 (b) The VP6 shell: 260 VP6 trimers are colored relativees—gold (T; on the threefold), red (T′; adjacent to T), darkfold), and purple (P′; adjacent to P). The inner VP2 layer is

Fig. 6. Alternative juxtapositions of VP6 trimers acrosslocal 2-fold axes in the T=13 icosahedral lattice. Theinterfaces between T and P′ and between T and D (blue)show a relative shift of ∼15 Å relative to those between Tand T and between T and T′ (tan). T–T has been aligned toT–D. Despite this shift, similar residues interact at thetrimer interface, generating a set of alternative salt bridges,as described in the text.


chain and main-chain polar groups from both VP6and VP2 and van der Waals contacts that includeLeu65, Thr68, Thr69, and Leu71 on one or both ofthe two VP6 monomers. Analysis of virus-likeparticles generated with wild-type VP2 shows thatany VP6 double mutant that contains a polar residueat Leu71 and a polar residue at either Leu70 orLeu65 fails to incorporate stably.11

Fig. 7. Interactions of VP6 and VP2. (a) View of the VP2 l(green) residues that have direct contacts with the VP6 trimerVP2 (side chains shown) receives a loop from VP6 (residues 65several conformations, depending on the part of the VP2 surfaadopts when interacting with the hydrophobic pocket shown

Only VP6-T has three identical contacts with VP2,and we can therefore expect it to be the most firmlyattached. Indeed, the overall thermal parameter (B-factor) for VP6 is lowest for VP6-T and increases fortrimers near the fivefold (Fig. 8). There is a similargradient in the BTV core.3 The β2–α3 loop of VP6-Pis displaced from VP2, leaving a “tunnel” beneaththe ring of VP6-P trimers, which are retained in theDLP only by VP6–VP6 contacts. In rotavirus strainWa, these peripentadic trimers dissociate fromtranscriptionally competent DLPs.12 It is plausibleto suppose that assembly of the VP6 layer on thesurface of VP2 proceeds from the threefold, ratherlike a two-dimensional crystallization. Some of theVP6–VP6 contacts in the DLP also appear to bepresent in tubular VP6 assemblies, an alternativesurface lattice of VP6 trimers.13

Internal structures

The capsid defined by residues 81–880 of VP2encloses a volume of 58×106 Å3, which must containthe 11 viral RNA segments, 11–12 copies each of VP1and VP3, and the N-terminal parts of the VP2subunits. Concentrically layered shells of density,with no regular contributions from higher-resolu-tion terms, occupy a substantial part of this volume(Fig. 9a). As in the electron density maps fromcrystals of reovirus and BTV cores,2,3,14 these shellsarise from the close packing of tightly coiled RNA.They do not extend under the 5-fold positions;instead, there is a well-defined, roughly cylindricalfeature projecting inward around each 5-fold axis,with rod-like substructures, some of which canclearly be interpreted as α-helices. We call thisstructure a “5-fold hub.” We attribute it to the N-terminal segments of the 10 abutting VP2 subunits,in agreement with a cryo-EM analysis of recombi-nant virus-like particles containing N-terminallydeleted VP2.15 We can assign at least one set of α-helices in the fivefold to residues 18–30, based on thepattern of identifiable side-chain density. Moreover,

ayer directly along the icosahedral threefold, highlightingbound over them. A hydrophobic pocket on the surface of–72). (b) Detailed view of this loop, which can adopt any ofce it contacts; it is represented here in the conformation itin (a).

Fig. 8. View of the DLP surface, colored by B-factor(low to high in green to red, respectively). The degree ofdisorder increases with radial position in the particle andwith distance within the surface from the nearesticosahedral 3-fold axis.


the entire hub is so situated that its outer rim extendsto the positions occupied by residue 81 of the fiveVP2A subunits and the likely position of thecorresponding residues of the five VP2B subunits(Fig. 9b). The overall volume occupied by the hub is∼2.5×105 Å3, roughly as expected for a compactlyfolded, hydrated structure composed of about 800amino acid residues (i.e., residues 1–80 of 10 VP2molecules). The same structure can be seen in thedensity map derived by single-particle reconstruc-tion from cryo-EM images of rotavirus DLPs.5

Discussion

We summarize our overall conclusions as follows.The VP6 trimer in the DLP has essentially the sameconformation as that in isolation.9 The bulk of theVP2 polypeptide chain (residues 81–880) folds into aplate-like structure resembling those of thecorresponding shell proteins, known as VP3 andλ1, in orbiviruses and orthoreoviruses, respectively.The two conformers of VP2, designated VP2A andVP2B, within the icosahedral asymmetric unit aresomewhat bent with respect to each other, butotherwise very similar. The N-terminal residues(residues 1–80) of VP2 form a previously undetect-ed, inward-projecting “fivefold hub,” condensedaround five radially directed α-helices. Although wecan assign a sequence within the hub only to thesehelices, it is clear that each hub contains five VP2AN-terminal arms and five VP2B N-terminal arms,which are known to be essential for recruiting VP1and VP3 into an assembling particle. Lower-resolu-tion RNA density is evident in three or fourconcentric layers in the regions between the five-

folds. We propose that the hubs function not only toorganize the assembly unit of the inner shell but alsoas spools around which genomic dsRNA segmentscan coil and as guides for extrusion of nascentmRNA.

Interpreting RNA density

When double-stranded nucleic acids are con-densed, either by ATP-driven packaging within aphage head or by polymer-induced compaction,the most stable configuration is a more or lessregular coil, with local close packing of adjacentsegments.16,17 An everyday analogy for the pack-ing of a stiff but bendable rod within a confinedregion is the winding of a garden hose in a hosepot. The energetics of nucleic acid condensationare dominated by electrostatic repulsion andresistance to bending. Both will cause the outer-most coils to pack tightly against the confiningstructure (in this case, the VP2 shell), andsuccessive layers will pack against the first. Themean spacing between adjacent segments of RNAdouble helix will be determined by the balance ofelectrostatic repulsion, which will space adjacentsegments as far from each other as possible, andstiffness, which will cause the RNA to avoid atight curvature (e.g., immediately adjacent to thespool). In double-stranded-DNA-containing phageheads, the former dominates; in underfilled heads,the DNA coils expand to fill the available internalvolume.17 In the absence of any specific grooves orstrongly varying patches of charge on the inward-facing surface of VP2, the outer layer of RNAdouble helices in rotavirus can be expected tocreate a uniform shell, except where the fivefoldhub (and associated VP1 and VP3) intervenes.The 11 dsRNA segments within a rotavirus

particle must coil independently because transcrip-tion of each appears to proceed independently, andthere are multiple rounds. Transcription is conser-vative: mRNA for each segment is copied by anassociated VP1 from the minus-sense templatestrand. The crowding of the viral interior providesinsufficient space for diffusion of VP1 and thereforerequires that each genomic segment spool throughits polymerase, with a propagating transcription“bubble” allowing the template strand to passthrough the protein. The nascent transcript exits atthe 5-fold axis,18 possibly guided by the fivefoldhub, with which the polymerase probably remainsat least loosely associated (see the text below). VP1must also generate the minus strand of eachsegment (and hence the genomic duplex segment)within a newly assembled particle, and the replica-tion mechanism may help organize the independentcoils. VP1 has a site to bind the 5′ 7-MeG cap on theplus-sense strand; this attachment may keep eachpolymerase molecule associated with its “own”genomic segment.19 Figure 11 illustrates the picturemost consistent with these constraints and with thedensity layers shown in Fig. 9a. A related picture hasbeen adduced for orbiviruses.14

Fig. 9. Internal features of the electron density map and their interpretation. (a) Contoured map in the vicinity of a 5-fold axis, showing layers of density (magenta) attributed to coiled RNA. Scale bar shows radius (distance from center) inangstroms. (b) The fivefold hub. Density from a 2Fo−Fc map, filtered to 7 Å resolution and contoured at 0.9σ, in the regionaround the 5-fold axis boxed in (a). (c) Schematic diagram of the interior structures. The relative spatial relationship ofthese features is shown, with suggested connections (dotted lines) indicating that the N-terminal extension of VP2 mayform an internal channel.


At low resolution, a dsRNA segment can beapproximated as a grooved cylinder with a 20-Å vander Waals diameter. Local hexagonal packing of RNAsegments, with packing diameter D (N20 Å, due toelectrostatic repulsion and hydration), will give riseto density maxima spaced equally inward at 0.87D(Fig. 10). The first such maximum will be located at adistance from the inward-facing surface of VP2 (whichis roughly spherical at the resolution set by D)determined by the van der Waals radius of an RNAcylinder and the thickness of the hydration layerbetween the protein and the RNA. The inward-facingradius of the VP2 shell (estimating the mean van derWaals surface from the atomic model) is 240±5 Å. The

radii of the three clearly defined RNA layers (Fig. 9)are 225 Å, 196 Å, and 167 Å. From these numbers, wecan estimate that D=33 Å. The volume within thecavity defined by the inner surface of VP2(5.8×107 Å3) must contain 11–12 copies each of VP1and VP3, the 12 fivefold hubs (residues 1–80 from 120copies of VP2), and the RNA. Assuming a partialspecific volume of 0.72 cm3/g for protein and anaverage hydration (water-accessible surface grooves,etc.) of 0.35 g/g, the RNA can then fill a total of4.8×107 Å3. A center-to-center distance ofD=33 Å in alocally hexagonal array and a linear density of 1 bpper 2.8 Å (A-form RNA) imply a volume of2.6×103 Å3/bp and, hence, a total RNA content of

Fig. 10. Schematic cross section of the rotavirus interior illustrating dsRNA packing. The diagram is drawnapproximately to scale (for an internal radius of ∼240 Å for the VP2 shell and for D; packing diameter of the dsRNAsegments, ∼33 Å). The dsRNA should be pictured as coiled around each fivefold, with approximate hexagonal closepacking in cross section. The principal density of the RNA helices is about 20 Å in diameter; the lighter region around eachcross section shows the hydration layer.


∼18.5 kbp, in excellent agreement with the totalnumber of base pairs in the rotavirus genome.We emphasize that the relatively well-ordered

organization of the genomic RNA does not dependon any specific interactions with VP2. It is simply theresult of tight packing and mechanisms for packag-ing, minus-strand synthesis, and multiple rounds oftranscription that avoid entanglement. Indeed, anyspecific or preferential interactions with particularsurfaces of VP2 would probably hinder polymeraseactivity by creating local “sticking points” for thedsRNA segments passing across them.

Fig. 11. Model for DLP assembly. VP2, blue (dark and lightRNA segments, magenta. VP1 recruits plus-strand RNA anddimers to form the fundamental assembly unit. Twelve suchVP3) come together to form a DLP; synthesis of minus-stramechanism by which one of each RNA segment is incorporat

The 5-fold hub

A clear definition of density for the fivefold hub,which was not detected in previous structuralanalyses of dsRNA virus particles, accounts for theN-termini of all VP2 subunits. We suggest that thehub has three distinct but related functions inrotavirus assembly, replication, and transcription.The first function is to organize the assembly unit ofthe inner shell, by recruiting a copy of VP1 or a copyof a VP1–VP3 complex, perhaps together with asegment of plus-strand RNA. The second function is

for A and B conformers, as in Fig. 2); VP1, red; VP3, orange;VP3, and these components come together with five VP2units (one lacking RNA and potentially lacking VP1 andnd RNA inside the particle reels in the plus strand. Theed remains unspecified.


to serve as a shaft around which coils the dsRNAsegment generated from the packaged plus strandby in situminus-strand synthesis. The third functionis to guide nascent mRNA out through its exitchannel (and perhaps to guide plus-strand RNA as itis pulled into the assembling core through minus-strand synthesis) (see Fig. 11).A VP2 decamer, organized by the fivefold hub and

associated with a VP1–VP3 complex, is a likelyassembly unit for the virion inner layer (Fig. 11).Although the hub is not required to form a VP2 shell[VP2 assembles into “single-layered particles” (SLPs)of correct size and shape, even when it lacks the firstN residues], it is essential for recruiting coexpressedVP1 or VP3.20 The proportion of incorporated VP1 isapproximately one molecule per decamer. A similarVP1/VP2 ratio is also optimal for polymeraseactivation. VP1 recognizes a conserved sequence atthe 3′ end of each viral plus-strand RNA segment, butinitiation of minus-strand synthesis requires VP2(with intact N-terminus) in a roughly 10:1 ratio.19,21

The hub thus appears to link the VP1/RNA complex(presumably with an associated VP3) and the shell-forming domains of VP2. The molecular character-istics of this linkage remain to be determined, butonly when the linkage has been established will VP1commence replication. Some association betweenVP1 and VP2 presumably persists even duringtranscription, as the polymerase must direct itstranscripts into the exit pore in the VP2 shell. Thefivefold hub has a channel along its axis, and apolymerase tethered near the tip of this channel couldreadily extrude transcripts along it.The layers of RNA density surround the fivefold

hubs. If the dsRNA is wound into independent coils,then the hubs are likely to be spools around whichthe RNA winds. It is possible that the way in whichdsRNA emerges from the polymerase helps togenerate these coils.19

RNA packaging and dsRNA synthesis

The properties of replication intermediates isolat-ed from infected cells indicate that assembly of VP1and VP2 into SLPs (and even into DLPs) precedesminus-strand synthesis, and that the plus-strandtemplate recruited by VP1 moves into the particle assynthesis proceeds.22 Figure 11 is a diagram of theRNA packaging process suggested by these observa-tions and by the DLP structure. In this model,progressiveminus-strand synthesis and formation ofa double-stranded product reel the plus strand intothe particle (SLP or DLP) and spool each segment ofthe genomic dsRNA around its fivefold hub. Certainaspects of Fig. 11 have formal similarities to othernucleic acid packaging processes. For example, inP22 and some other bacteriophage heads, thereappears to be a protein shaft around which theDNA can coil as it enters,23 driven in that case by anATP-dependent pump at the portal (rather than bythe activity of an internal polymerase).Various features of the mechanism embodied in

Fig. 11 may not apply to other dsRNA viruses. For

example, in dsRNA phage such as ϕ6, the plus-strand RNA templates are fully packaged beforedsRNA synthesis starts.24 The orthoreovirus poly-merase λ3 does not seem to require the core shellprotein λ1 to initiate polymerization, nor havestructural studies suggested recognition of 3′sequences by the λ3 template channel,25 but it ispossible that slightly different local mechanismslead to the same larger-scale processes.A remaining puzzle posed by dsRNA virus

assembly is the mechanism that selects for packagingone copy each of the various gene segments. If apicture such as the one embodied in Fig. 11 is correct,some sort of sequence-specific and structure-specificinteraction among the plus-strand segments is likelyto determine selectivity; viral nonstructural proteins(e.g., NSP2)26,27 might also contribute. The interac-tions and structures determining RNA selectionmight occur only outside the synthetically activeparticle, and they would then dissolve duringcompletion of dsRNA synthesis and RNA packaging.

Materials and Methods

Purification and crystallization

Rotavirus (UK Bovine isolate) DLPs were prepared inthe laboratory of A. R. Bellamy (University of Auckland) inaccordance with the procedure described by Street et al.28

In brief, virus was propagated in MA104 cells grown in1585-cm2 roller bottles. Cells were harvested after 70 h, andthe DLPs were purified from the lysate of a concentratedcell pellet. We obtained identical crystals from DLPsprepared in this way (i.e., particles are isolated beforethey have budded into the endoplasmic reticulum andmatured into virions) and from DLPs prepared byuncoating purified virions. The particles were transportedin CsCl2 as recovered from the final CsCl2 gradient.Following dialysis against 10 mMHepes (pH 7.5), 150 mMNaCl, and 0.02% sodium azide, the particles werecentrifuged at 41,000 rpm in a Beckman SW41 swingingbucket rotor for 2 h, and the pellet was resuspended in thedialysis buffer to a concentration of 25 mg/mL and useddirectly for crystallization.The DLP crystallized using the hanging-drop method

with drops containing 1 μL of protein and 1 μL of wellsolution at 22 °C. The optimized well solution contained50 mM NaCl, 375 mM Na2SO4, 1.6% polyethylene glycol10,000, 0.02% sodium azide, and 20 mM Hepes (pH 7.5).Crystals appeared within 3 weeks and grew tomaximum size (0.5 mm×0.5 mm×0.4 mm) in ∼6 months.Initial X-ray analysis determined that the crystals wereof space group P212121 with the following unit celldimensions: a=740.75 Å, b=1198.07 Å, and c=1345.41 Å.There were four rotavirus DLPs per unit cell, with eachcrystallographic asymmetric unit containing an entireparticle.

Data collection and processing

Crystals were gently transferred to a stabilizationsolution containing a well solution with 8.5% β-D-octylglucopyranoside to reduce surface tension. Crystalswere mounted in quartz capillaries and stored until

Table 1. Crystallographic statistics

Data collection statisticsResolution limit (Å) 223–3.8Space group P212121Unique reflections 2,393,850Redundancya 1.1 (1.0)Completenessa (%) 20.8 (8.2)I/σa 3.4 (1.0)Rsym

a,b (%) 18.2

Refinement statisticsReflections used in refinement 1,823,596Polypeptide chains 900Protein atoms 3,233,140Residues in allowed regions of

Ramachandran plot VP2 (%)97.2

Residues in the most favored regionsof Ramachandran plot VP2 (%)

60.6

Residues in allowed regions ofRamachandran plot VP6 (%)

99.7

Residues in the most favored regionsof Ramachandran plot VP6 (%)

78.1

rmsd bond lengths (Å) 0.011rmsd bond angles (°) 1.67Mean B-value (Å2) 164.0Resolution range (Å) 30–3.8R-factora,c 32.8 (55.9)Free R-factora,c 33.9 (57.9)

a Values for the last shell are given in parentheses.b Rsym=∑(I− hIi)/∑ I, where hIi is the average intensity over

symmetry-equivalent reflections.c R-factor=∑(Fobs−Fcalc)/∑Fobs, where the summation is over

the working set of reflections. For the free R-factor, the summationis over the test set of reflections (2% of the measured reflectionsselected in thin shells).


needed. Diffraction data at Bragg spacings between 215 Åand 3.8 Å were collected at CHESS beamline F1 using adual-plate Quantum 4 CCD detector at a distance of750 mm with 0.1° oscillations at a single wavelength. Thebeam was collimated to 100 μm, and cross fire was furtherreduced using hutch entrance slits, which were about 2 mupstream of the defining collimator. Each crystal wasexposed (40 s) once for every crystal volume, as defined bybeam size. Only data from crystals with a mosaicity ofb0.055° (over 2.3 million reflections) were used for thefinal structure determination. Reflections were indexedand integrated with a modified version of Denzo29 andscaled using SCALA and POSTREF.30 Statistics are shownin Table 1. The data were 20.8% complete overall, with8.2% completeness in the final 4.0–3.8 Å shell. Sixty-foldredundancy implies that there is some phase extensionpower, even at the limiting resolution.

Initial map calculation

Starting phases at 18.2 Å were obtained from a cryo-EM reconstruction of the DLP at ∼20 Å (kindlyprovided by B. V. V. Prasad, Baylor College ofMedicine). The map was modified to that as describedby Reinisch et al., so that it had a uniform positivedensity within a defined contour level and zero densityoutside.2 The initial orientation of the electron micros-copy reconstruction within the unit cell was determinedfrom an icosahedral locked self-rotation function31 andtranslation function (AMORE)32 using structure factorsfrom the modified map. During the course of phaseextension, orientation parameters were optimized by

monitoring improvements to Fo/Fc correlations whenapplying small shifts to the particle rotation/translation.

Phase extension

Phase extension proceeded in steps of one reciprocallattice point, using RAVE and associated programsessentially as described by Reinisch et al.2 Fcalc from theaveraged map was used for missing Fobs. At each cycle,the density map was averaged in a P1 cell that covereda single DLP, and the averaged density was theninterpolated back into the crystallographic unit cell forphase generation to begin the next cycle. The initialaveraging mask was generated using the outer surfaceof the starting 18.2-Å model with an inner radius of∼200 Å. The mask was periodically recalculated athigher resolution on a finer grid, as required usingCCP4 (EXTEND)30 and RAVE (MAPMAN and MAMA)routines.33

Model building and refinement

The final m(2Fo−Fc) (where m is the figure of merit)map from the phase extension procedure was sharp-ened with a B-factor of −150. It showed clear densityfor the 13 (per icosahedral asymmetric unit) VP6subunits and the 2 VP2 monomers per icosahedralasymmetric unit, for which no structure had previouslybeen determined. There was connected density for theVP2 main chain and reasonable density for many sidechains, and the VP2 polypeptide chain could be builtwithout great difficulty into the density with theprogram O.34 Parts of the VP6 monomers, especiallythose at the VP6–VP2 interface, were also rebuilt withreference to the density. The model was refined using aversion of CNS35 specifically compiled to take advan-tage of the additional memory available on a 64-bitOpteron processor. Positional refinement, restrainedgroup B-factor refinement, and difference map genera-tion in CNS proceeded using 60-fold NCS constraintswithin the icosahedral asymmetric unit and additionalsets of NCS restraints among the 13 VP6 monomers andthe 2 VP2 monomers. Lower-resolution 2mFo−DFcmaps were also calculated to visualize the interiordensities corresponding to both the RNA layers and thefeatures about the fivefold.The final refined model includes residues 81–880 of

VP2 and residues 1–397 of VP6, with Rwork and Rfree of34.1 and 34.9, respectively (Table 1). The high non-crystallographic symmetry diminishes the cross-valida-tion power of Rfree; we suggest that lack of divergencefrom the working R is nonetheless a meaningfulindication that we have avoided overrefinement in theoutermost shells, where data completeness correspondsto a much lower effective redundancy (see the textabove). Overall, 91.1% of the residues in VP2 fall in themost favored and additionally allowed regions of theRamachandran plot; the generously allowed regioncontains 6.1%, and the disallowed region contains theremaining 2.7%. The rmsd from ideal bond lengths is0.010 Å and that from ideal angles is 1.7°.

Accession numbers

Coordinates have been deposited in the Protein DataBank with accession number 3KZ4.


Acknowledgments

We thank I. Anthony and S. Greig for assistancewith virus propagation and purification, BarbaraHarris for advice and assistance with crystallization,and the staff of CHESS and MacCHESS for cheerfulsupport. The research was supported by NationalInstitutes of Health grant CA13202 (to S.C.H.) andby project grants from the Health Research Councilof New Zealand (to A.R.B.). S.C.H. is an Investigatorat the Howard Hughes Medical Institute.

References

1. Schiff, L. A., Nibert, M. L. & Tyler, K. L. (2007).Orthoreoviruses and their replication. In (Knipe, D.M. & Howley, P. M., eds), pp. 1853–1915, Lippincott,Williams and Wilkins, Philadelphia, PA.

2. Reinisch, K. M., Nibert, M. L. & Harrison, S. C. (2000).Structure of the reovirus core at 3.6 Å resolution.Nature, 404, 960–967.

3. Grimes, J. M., Burroughs, J. N., Gouet, P., Diprose, J. M.,Malby, R., Zientara, S. et al. (1998). The atomic structureof the bluetongue virus core. Nature, 395, 470–478.

4. Tihova, M., Dryden, K. A., Bellamy, A. R., Greenberg,H. B. & Yeager, M. (2001). Localization of membranepermeabilization and receptor binding sites on theVP4 hemagglutinin of rotavirus: implications for cellentry. J. Mol. Biol. 314, 985–992.

5. Zhang, X., Settembre, E., Xu, C., Dormitzer, P. R.,Bellamy, R., Harrison, S. C. & Grigorieff, N. (2008).Near-atomic resolution using electron cryomicro-scopy and single-particle reconstruction. Proc. NatlAcad. Sci. USA, 105, 1867–1872.

6. Yu, X., Jin, L. & Zhou, Z. H. (2008). 3.88 Å structure ofcytoplasmic polyhedrosis virus by cryo-electron mi-croscopy. Nature, 453, 415–419.

7. Li, Z., Baker, M. L., Jiang, W., Estes, M. K. & Prasad,B. V. (2009). Rotavirus architecture at subnanometerresolution. J. Virol. 83, 1754–1766.

8. Coulibaly, F., Chevalier, C., Gutsche, I., Pous, J.,Navaza, J., Bressanelli, S. et al. (2005). The birnaviruscrystal structure reveals structural relationshipsamong icosahedral viruses. Cell, 120, 761–772.

9. Mathieu, M., Petitpas, I., Navaza, J., Lepault, J.,Kohli, E., Pothier, P. et al. (2001). Atomic structure ofthe major capsid protein of rotavirus: implicationsfor the architecture of the virion. EMBO J. 20,1485–1497.

10. Heiman, E. M., McDonald, S. M., Barro, M., Tarapor-ewala, Z. F., Bar-Magen, T. & Patton, J. T. (2008).Group A human rotavirus genomics: evidence thatgene constellations are influenced by viral proteininteractions. J. Virol. 82, 11106–11116.

11. Charpilienne, A., Lepault, J., Rey, F. A. & Cohen, J.(2002). Identification of rotavirus VP6 residues locatedat the interface with VP2 that are essential for capsidassembly and transcriptase activity. J. Virol. 76,7822–7831.

12. Greig, S. L., Berriman, J. A., O'Brien, J. A., Taylor, J. A.,Bellamy, A. R., Yeager, M. J. & Mitra, A. K. (2006).Structural determinants of rotavirus subgroup speci-ficity mapped by cryo-electron microscopy. J. Mol.Biol. 356, 209–221.

13. Lepault, J., Petitpas, I., Erk, I., Navaza, J., Bigot, D.,Dona, M. et al. (2001). Structural polymorphism of the

major capsid protein of rotavirus. EMBO J. 20,1498–1507.

14. Gouet, P., Diprose, J. M., Grimes, J. M., Malby, R.,Burroughs, J. N., Zientara, S. et al. (1999). The highlyordered double-stranded RNA genome of blue-tongue virus revealed by crystallography. Cell, 97,481–490.

15. Lawton, J. A., Zeng, C. Q., Mukherjee, S. K., Cohen, J.,Estes, M. K. & Prasad, B. V. (1997). Three-dimensionalstructural analysis of recombinant rotavirus-likeparticles with intact and amino-terminal-deletedVP2: implications for the architecture of the VP2capsid layer. J. Virol. 71, 7353–7360.

16. Maniatis, T., Venable, J. H., Jr. & Lerman, L. S. (1974).The structure of psi DNA. J. Mol. Biol. 84, 37–64.

17. Earnshaw, W. C. & Harrison, S. C. (1977). DNAarrangement in isometric phage heads. Nature, 268,598–602.

18. Lawton, J. A., Estes, M. K. & Prasad, B. V. (1997).Three-dimensional visualization of mRNA releasefrom actively transcribing rotavirus particles. Nat.Struct. Biol. 4, 118–121.

19. Lu, X., McDonald, S. M., Tortorici, M. A., Tao, Y. J.,Vasquez-Del Carpio, R., Nibert, M. L. et al. (2008).Mechanism for coordinated RNA packaging andgenome replication by rotavirus polymerase VP1.Structure, 16, 1678–1688.

20. Zeng, C. Q., Estes, M. K., Charpilienne, A. & Cohen, J.(1998). The N terminus of rotavirus VP2 is necessaryfor encapsidation of VP1 and VP3. J. Virol. 72, 201–208.

21. Gallegos, C. O. & Patton, J. T. (1989). Characterizationof rotavirus replication intermediates: a model for theassembly of single-shelled particles. Virology, 172,616–627.

22. Patton, J. T. & Gallegos, C. O. (1990). Rotavirus RNAreplication: single-stranded RNA extends from thereplicase particle. J. Gen. Virol. 71, 1087–1094.

23. Johnson, J. E. & Chiu, W. (2007). DNA packaging anddelivery machines in tailed bacteriophages. Curr.Opin. Struct. Biol. 17, 237–243.

24. Qiao, X., Qiao, J. & Mindich, L. (1997). Stoichiometricpackaging of the three genomic segments of double-stranded RNA bacteriophage phi6. Proc. Natl Acad.Sci. USA, 94, 4074–4079.

25. Tao, Y., Farsetta, D. L., Nibert, M. L. & Harrison, S. C.(2002). RNA synthesis in a cage—structural studies ofreovirus polymerase lambda3. Cell, 111, 733–745.

26. Jayaram, H., Taraporewala, Z., Patton, J. T. & Prasad,B. V. (2002). Rotavirus protein involved in genomereplication and packaging exhibits a HIT-like fold.Nature, 417, 311–315.

27. Jiang, X., Jayaram, H., Kumar, M., Ludtke, S. J., Estes,M. K. & Prasad, B. V. (2006). Cryoelectron microscopystructures of rotavirus NSP2–NSP5 and NSP2–RNAcomplexes: implications for genome replication.J. Virol. 80, 10829–10835.

28. Street, J. E., Croxson, M. C., Chadderton, W. F. &Bellamy, A. R. (1982). Sequence diversity of humanrotavirus strains investigated by Northern blot hy-bridization analysis. J. Virol. 43, 369–378.

29. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode.Methods Enzymol, 276, 307–326.

30. Collaborative Computational Project. (1994). TheCCP4 suite: programs for protein crystallography.Acta Crystallogr. Sect. D, 50, 760–763.

31. Tong, L. & Rossmann, M. G. (1997). Rotation functioncalculations with GLRF program. Methods Enzymol.276, 594–611.


32. Navaza, J. (2001). Implementation of molecular replace-ment in AMoRe. Acta Crystallogr. Sect. D, 57, 1367–1372.

33. Jones, T. A. (1992). CCP4 Proceedings. SERC DaresburyLaboratory, Warrington, UK.

34. Jones, T. A., Zou, J. Y., Cown, S. W. & Kjeldgaard, M.(1991). Improved methods for building proteinmodels in electron density maps and the location of

errors in these models. Acta Crystallogr. Sect. A, 47,110–119.

35. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography and NMR system: a new softwaresuite for macromolecular structure determination.Acta Crystallogr. Sect. D, 54, 905–921.

X-ray Crystal Structure of the Rotavirus Inner Capsid Particle at 3.8 Å ResolutionIntroductionResultsStructure determinationVP2VP6VP2–VP6 contactsInternal structures

DiscussionInterpreting RNA densityThe 5-fold hubRNA packaging and dsRNA synthesis

Materials and MethodsPurification and crystallizationData collection and processingInitial map calculationPhase extensionModel building and refinementAccession numbers

AcknowledgmentsReferences

Documents

X-ray Crystal Structure of the Rotavirus Inner Capsid Particle ...crystal.harvard.edu/wp-content/uploads/2018/12/mcclain.pdfof polypeptide chain (residues 81–100) leading into them