The structure of a birnavirus polymerase revealsa distinct active site topologyJunhua Pan*, Vikram N. Vakharia†, and Yizhi Jane Tao*‡

*Department of Biochemistry and Cell Biology, Rice University, Houston, TX 77005; and †Center for Biosystems Research, University of MarylandBiotechnology Institute, College Park, MD 20742

Edited by Michael G. Rossmann, Purdue University, West Lafayette, IN, and approved March 13, 2007 (received for review December 27, 2006)

Single-subunit polymerases are universally encoded in both cellu-lar organisms and viruses. Their three-dimensional structures havethe shape of a right-hand with the active site located in the palmregion, which has a topology similar to that of the RNA recognitionmotif (RRM) found in many RNA-binding proteins. Considering thatpolymerases have well conserved structures, it was surprising thatthe RNA-dependent RNA polymerases from birnaviruses, a groupof dsRNA viruses, have their catalytic motifs arranged in a per-muted order in sequence. Here we report the 2.5 Å structure of abirnavirus VP1 in which the polymerase palm subdomain adopts anew active site topology that has not been previously observed inother polymerases. In addition, the polymerase motif C of VP1 hasthe sequence of -ADN-, a highly unusual feature for RNA-depen-dent polymerases. Through site-directed mutagenesis, we haveshown that changing the VP1 motif C from -ADN- to -GDD- resultsin a mutant with an increased RNA synthesis activity. Our resultsindicate that the active site topology of VP1 may represent a newlydeveloped branch in polymerase evolution, and that birnavirusesmay have acquired the -ADN- mutation to control their growthrate.

evolution � virus � RdRp

S ingle-subunit polymerases, including RNA/DNA-dependentRNA/DNA polymerases, are universally encoded in cellular

organisms and viruses. Sequence analysis shows that RNA-dependent RNA polymerases (RdRps) are a cluster of closelyrelated enzymes that are mostly found in viruses, in which theyassume critical functional roles by replicating and transcribing viralgenome. The recently reported crystal structures of several RdRpsand their functional complexes have provided important insightsinto the biological functions and catalytic mechanisms of theseenzymes (1–10). These structures contain a well conserved corepolymerase domain, the shape of which resembles a right hand withfingers, palm, and thumb subdomains. The palm, which is the mostconserved region, contains a central, nonvariant, four-stranded�-sheet with five recurring catalytic sequence motifs (from A to E).

In contrast to these conventional features, birnavirus polymeraseVP1 exhibits several unusual characteristics (11, 12). Most notice-ably, the essential -X(G)DD- sequence, which is often referred toas the polymerase motif C (13, 14), is missing from VP1 (12). Thisraises the question as to whether VP1 promotes catalysis by thetwo-metal mechanism like in conventional polymerases, or whetherVP1 employs a different mechanism for nucleotidyl transfer. Re-cent results indicate that birnavirus VP1, as well as the polymerasesfrom some tetraviruses, may belong to a special group of uncon-ventional polymerases with five essential RNA polymerase motifsarranged in the permuted order of C–A–B–D–E (15). In addition,the motif C in birnaviruses may have the sequence -ADN-, resultingin only two aspartate residues in the active site (15).

Like polymerases from other dsRNA viruses, birnavirus VP1catalyzes both replication and transcription of the bisegmented viralgenome. Intact birnavirus particles are transcription-competent,and they are capable of producing viral messengers in a semicon-servative fashion (16). In vitro, isolated VP1 proteins have beenshown to possess a replicase activity using viral (�)RNA templates

(17). Three-dimensional structures of birnaviruses show that theordered capsid shell is made of VP2, whereas the viral genome,polymerase VP1, and VP3, another major capsid protein, areinternally disordered (18–20). The VP2 capsid shell has a T � 13icosahedral symmetry, but the ‘‘T � 2’’ viral core, which is com-monly observed in dsRNA viruses, is missing in birnaviruses.

Birnavirus polymerase VP1 initiates RNA synthesis via protein-priming (21–24). Protein-priming is an important mechanism thatmany viruses (e.g., adenovirus, picornavirus, bacteriophage �29)use to initiate genome replication, thereby preventing the loss ofterminal sequence information (25). Although the protein primerand the polymerase are usually two separate molecular moieties,the protein priming function in birnaviruses is carried out by theRNA-dependent RNA polymerase VP1 itself. It has been shownthat both virion-associated and recombinant birnavirus VP1 havethe self-guanylylation activity (12, 17, 22, 26). VP1 self-guanylylation, which does not require viral RNA template, pro-duces two products, VP1-pG and VP1-pGpG (12, 17, 22, 26). It hasbeen proposed that the pGpG moiety in VP1-pGpG binds to theconserved pCpC sequence at the terminal end of the viral RNAtemplate during the initiation of nucleotide polymerization (17, 27).Consequently, the 5� ends of both genomic and messenger RNAsof birnaviruses are covalently linked to a VP1 molecule.

To elucidate whether birnavirus VP1 is a permuted polymeraseand how it catalyzes template-independent protein priming, wehave determined the crystal structure of a birnavirus polymeraseVP1 from the infectious bursal disease virus (IBDV), a well studiedbirnavirus that causes severe immunosuppression in avian species.The structure reveals several highly unusual features. First, IBDVVP1 adopts a unique active site topology, which brings the fiveessential RNA polymerase motifs in the permuted order of C–A–B–D–E to form a conserved catalytic active site. Second, the-GDD- motif strictly conserved in many other polymerases isindeed replaced by -ADN- in IBDV VP1. Converting the sequencemotif -ADN- to -GDD- by site-directed mutagenesis resulted in amutant with an increased polymerase activity. Third, the putativeguanlylylation site Ser-166 (17) is located �23 Å from the poly-merase active site. In addition, RNA modeling on VP1 suggests thatterminal protein-priming by VP1 would require large movements ofseveral structural modules.

Results and DiscussionBiochemical Characterization of IBDV VP1. Isolated IBDV VP1 existsas monomers in solution, as evidenced by gel filtration chromatog-

Author contributions: Y.J.T. and V.N.V. designed research; J.P. and Y.J.T. performed re-search; V.N.V. contributed new reagents/analytic tools; J.P. and Y.J.T. analyzed data; andJ.P. and Y.J.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Abbreviation: IBDV, infectious bursal disease virus.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.pdb.org (PDB ID code 2PGG).

‡To whom correspondence should be addressed. E-mail: ytao@rice.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0611599104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0611599104 PNAS � May 1, 2007 � vol. 104 � no. 18 � 7385–7390

BIO

CHEM

ISTR

Y

raphy and electron microscopy [supporting information (SI) Fig.5a]. Recombinant IBDV VP1 exhibits a self-guanylylation activityin vitro and becomes radio-labeled in the presence of [�-32P]rGTP(SI Fig. 5b). Guanylylation of IBDV VP1 does not require a viralRNA template (SI Fig. 5b, lane 1). RNA replication activities wereobserved in the presence of a virus-specific RNA, but not withnonspecific ssDNA molecules (SI Fig. 5b, lanes 2 and 3). Theseresults are similar to the activities previously reported for VP1 froma homologous birnavirus, the infectious pancreatic necrosis virus(26, 28). Using limited proteolysis, we have identified a stable corefragment VP1c that encompasses amino acid residues 19–810.VP1c forms monomers in solution and retains the self-guanylylation and polymerase activities, similar to the full-lengthprotein.

Overview of the Structure. The crystal structure of IBDV VP1c wasdetermined to 2.5Å resolution using multiple isomorphous replace-ment and anomalous scattering (MIRAS). VP1c is an oval-shapedmolecule that is �50 � 50 � 60 Å3 in size. The final model con-tains amino acid residues from 31 to 602 and 612 to 804 (Fig. 1a andSI Fig. 6). The polypeptide can be divided into three functionalregions, namely the central polymerase domain (amino acids 168–658), the N-terminal (amino acids 1–167) and the C-terminal(amino acids 659–878) domains (Fig. 1). The central polymerasedomain folds into a structure of a right-hand shape ( fingers–palm–thumb) with a central canyon running through the whole molecule.Both N- and C-terminal domains have an extended conformation,and they span the canyon at the back and front of the polymerasedomain, respectively.

The palm of IBDV VP1 Has an Unusual Active Site Topology. Closeinspection of VP1 structure reveals that its palm has adopted atopology that has not been observed in any other RNA or DNApolymerases. The �-hairpin, which is formed by secondary structureelements P�1 and P�2 and contains the polymerase motif C, isconnected to P�1 and P�3 in VP1 (Fig. 2). The VP1 polymeraseactive site is formed by the motif C in the �-hairpin and motif Afrom the neighboring �-strand P�3. However, in a conventionalRNA polymerase, P�1 and P�3 would be directly connected, andthe �-hairpin containing motif C would be inserted between P�2and P�3, which are located in the palm II region (Fig. 2). Forexample, the �-hairpin P12 and P13 from reovirus polymerase �3,which is equivalent to P�1 and P�2 in VP1, is inserted betweenP�13 and P�16 that are equivalent to P�2 and P�3 in VP1 (Fig. 2).At the amino acid sequence level, it is evident that motif C in VP1

has been relocated from its conserved site, as shown in reovirus �3,to an upstream position immediately in front of motif A by �120aa residues (Fig. 2). The topology of IBDV VP1 seen in ourstructure closely matches the prediction made by Gorbalenya et al.(15), who proposed that birnavirus, as well as a small number oftetraviruses, may have a permuted polymerase fold based onsequence analysis.

Spatial Positions of The Five Polymerase Motifs Are Conserved in IBDVVP1. Except for the different connectivity involving the �-hairpin,the rest of the palm resembles that of other viral RNA polymeraseswith no large-scale deletions or insertions of secondary structureelements. In particular, the four other polymerase motifs, includingA, B, D, and E, are arranged in the same sequential order as inconventional polymerases. The polymerase motif B folds into astrand-turn-helix structure at the interface between the fingers IIand the palm II regions (Figs. 1–3). In nearly all RdRps, motif B hasthe consensus sequence of -SGxxxT- and has been proposed tointeract with the 2� OH group on the incoming nucleotide (1, 7).

Fig. 1. IBDV VP1c crystal structure. (a) The ribbon diagram. The N- and C-terminal domains are colored in yellow and magenta, respectively. The centralpolymerase domain is shown in three different colors with the fingers in blue, the palm in red, and the thumb in green. An internal disordered region (aminoacids 603–611) is shown by a dashed line. The molecule is viewed from the downstream end of the active site canyon. (b) Secondary structure assignment of IBDVVP1. Disordered regions are shown by dashed lines, whereas cylinders and arrows represent �-helices and �-strands, respectively. Terminal regions removed bychymotrypsin are not assigned with any symbols. Conserved polymerase motifs are highlighted in gray in the order of G, F, C, A, B, D, and E.

Fig. 2. New topology adopted by IBDV VP1 palm. (a) Domain assignmentwith polypeptide regions colored according to Fig. 1a. F1, fingers region I; P1,palm region I; F2, fingers region II, P2, palm region II; T, thumb. The two palmregions are colored in red and pink, respectively, for clarity. Motif C, high-lighted by a small box, is always located in the palm region II in conventionalpolymerases (e.g., reovirus �3), but is now observed in the palm region I inIBDV VP1. (b) IBDV VP1 palm. (c) The palm of reovirus �3. The �-hairpincontaining motif C is connected to non-equivalent secondary structure ele-ments in IBDV VP1 and reovirus �3, resulting in different topologies.

7386 � www.pnas.org�cgi�doi�10.1073�pnas.0611599104 Pan et al.

Motif B may also function in guiding the template entry into theactive site, as it has been shown to form the base of the template-entry channel (29). The polymerase motif D in IBDV VP1 formsa helix–turn–strand structure (Figs. 1 and 3). The �-strand of thisstructure forms the last strand in the central four-stranded �-sheetin the palm. The conserved lysine residue K529 is likely to interactwith nucleotide substrate to facilitate its entry into the active site(Fig. 4). The last �-hairpin in the palm, consisting of P�6 and P�7,forms the polymerase motif E (Figs. 1 c and d and 3), which is alsoan important part of the primer grip, as it has been shown to interactwith the primer strand. Amino acid residues from motif E aremostly hydrophobic but often poorly conserved (Fig. 3). As in otherRNA-dependent polymerases, the primer grip forms a second�-sheet with the �-strand (T�1) at the N terminus of the thumb.Motif E may have an important function in the proper positioningof the thumb relative to the palm.

Structure comparison indicates that the five recurring polymer-ase motifs from IBDV VP1 adopt similar secondary structures andoccupy nearly identical spatial locations to those observed in otherviral RNA polymerases. Superposition of 61 C� atoms from the fivepolymerase motifs resulted in RMS distance deviations of 1.8–3.0Å (Fig. 3 and Table 1). Among the eight polymerases used forcomparison, reovirus �3, HCV polymerase, and HIV-RT gave

deviations �2 Å. The close match of these motifs from VP1 to otherpolymerases indicates that by adopting the new topology, IBDVVP1 is able to accommodate the dramatic shuffling of catalyticmotifs in sequence without changing the overall 3D structure andactive site configuration.

Fingers, thumb, and the N- and C-Terminal Domains. The fingerssubdomain (formed by residues 168–360 and 420–489) consists of12 �-helices and nine �-strands. Many secondary structure ele-ments in VP1 can be matched to those found in HCV, reovirus, and�6 polymerases, although there are significant differences in theirsize and orientation. A twisted, four-stranded �-sheet (F�3, F�5,F�8, F�9) along with a 15-residue loop (residues 321–335) formsthe fingertip structure unique to RdRps (30). VP1 fingertip does notinteract directly with the thumb. The rNTP binding loop, oftenreferred to as the polymerase motif F (residues 333–339), is locatedin the fingertip region. Basic residues in the rNTP binding loop havebeen shown to interact with phosphates on the incoming nucleotide.A newly recognized polymerase motif G (15) (residues 253–363), aconsensus sequence of -SX1–2G- flanked by two lysine residues, isalso found in the VP1 fingers. Although sequence similaritiesbetween VP1 and Ras-type GTP binding motifs had suggested thatmotif G participated in VP1 self-guanylylation during proteinpriming (12), the crystal structure of VP1 shows a rather differentconformation and therefore does not support such a notion.

The thumb subdomain (residues 580–658) folds into a �-strandfollowed by four �-helices. The initial �-strand forms a small�-sheet, called the ‘‘primer grip,’’ with two other �-strands from thepalm (see previous discussions).

The N-terminal domain (residues 1–168) of IBDV VP1 folds intoa mixed �/� structure. Previous results have implicated VP1 N-terminal domain in protein priming as it possesses the putativeguanylylation site residue (see discussion below). The L-shapedN-terminal domain interacts with the fingers and thumb of the core

Fig. 3. IBDV VP1 structure compared with other viral polymerases. (a) Structure-based sequence alignment of the seven polymerase motifs. These polymerases areall RdRps from either dsRNA or ss(�)RNA viruses except for HIV-RT. The complete virus names are provided in Materials and Methods. Green, yellow, and blue highlightstrictly conserved, highly conserved, and generally conserved amino acid residues, respectively. (b) Superposition of IBDV VP1 and reovirus �3 active sites. The substrateCTP and divalent metal ions are from the reovirus �3 elongation complex (1).

Fig. 4. RNA modeling onto IBDV VP1. The dsRNA duplex is brought fromreovirus elongation complex by superimposing 61 conserved C� atoms, asdescribed in Table 1. The VP1 model in this figure contains residues 169–579.The thumb, N-terminal, and C-terminal domains have been removed from themolecule to achieve a better view of the polymerase active site. Importantamino acid residues are labeled. Molecular surface is colored according toelectrostatic potential. In the active site, the template is shown in red; nascent,green; metal ion, yellow spheres; nucleotide substrate, pale blue.

Table 1. Distance rmsd when different viral polymerase pairs aresuperimposed based on motifs A–E

rmsd, Å Reovirus Phi6 RHDV BVDV HCV Rhino Norwalk HIV-RT

IBDV 1.76 2.61 2.22 2.23 1.97 2.39 2.99 1.95Reovirus 2.25 1.43 1.67 1.56 1.80 2.56 2.17Phi6 2.13 2.29 2.24 2.41 2.89 2.68RHDV 1.51 1.36 1.34 2.07 2.20BVDV 0.98 1.71 2.50 2.00HCV 1.73 3.03 1.92Rhino 1.82 2.42Norwalk 2.83

Numbers are for a total of 61 C� atoms only.

Pan et al. PNAS � May 1, 2007 � vol. 104 � no. 18 � 7387

BIO

CHEM

ISTR

Y

polymerase with its short and long arms, respectively, to maintainthe VP1 active site in a closed conformation. The C-terminaldomain of the VP1 structure (residues 659–804) is mostly �-helical.It starts at the thumb, runs across the canyon in the front of thepalm, and then wraps around the fingers subdomain.

The Active Site and Modeling of the Elongation Complex. Superpo-sition of the VP1 crystal structure with structures of the catalyticcomplexes of reovirus �3 (1), the closest structural homolog of VP1(Fig. 3 and Table 1), suggests that the VP1 active site has a ‘‘closed’’conformation even in the absence of a bound substrate. To provideinsights into VP1 RNA binding and catalysis, we have modeled thereovirus �3 elongation complex, including a short RNA duplex, anNTP substrate, and two Mn2� ions, onto the VP1 structure. TheRNA duplex and the NTP substrate fit into the VP1 catalytic cavitywith very few steric clashes (Figs. 3b and 4).

The most noticeable feature of the VP1 active site is thesubstitution of a strictly conserved aspartate residue by an aspar-agine at position 403 (Fig. 3b). Therefore, VP1 has only twoaspartates in its active site, as compared with three found inpolymerases from nearly all dsRNA and �ssRNA viruses. Never-theless, the three amino acids D402, N403 and D416 assume nearlyidentical positions and conformations as the three aspartic acidresidues D734, D735, and D585, respectively, from reovirus �3.Inspection of the VP1 structure shows that there are no otheraspartate or glutamate residues within at least 6Å from N403. The-534IDD- sequence, originally identified as motif C in IBDV VP1(12), is found in the last �-strand of the central four-stranded�-sheet in the palm. Situated at the interface between the palm andthe thumb, -534IDD- is far from the active site and should notdirectly participate in catalysis.

Previous studies have shown that the major functional role of thethree highly conserved aspartates is to coordinate a pair of metalions (31). Two of the aspartates, including the aspartate from motifA (equivalent to Asp-416 in VP1) and the first aspartate from motifC (equivalent to Asp-402 in VP1), are absolutely essential forpolymerase function (32–34). The second aspartate from motif C(equivalent to Asn-403 in VP1) is also required for the properfunctioning of RNA polymerases. For example, the Asp 3 Asnmutation in a calicivirus polymerase resulted in a complete loss ofactivity (32), and the same change in HCV NS5B produced amutant with low levels of polymerase activity in vitro (33). Addi-tionally, an Asp3 Asn mutation would make poliovirus polymer-ase inactive, although the mutant can be rescued by substitutingMg2� with Mn2� for metal cofactors (34).

Birnavirus VP1 of both virion-associated and recombinant formscan use either Mg2� or Mn2� as the metal cofactor for catalysis (17,35), unlike the above-mentioned poliovirus polymerase mutant(34). Therefore, the question arises as to how birnavirus VP1efficiently binds to Mg2� ions under physiological conditions withonly two aspartic acid residues in the active site. It has been reportedthat the asparagines in protein molecules can undergo deamidationto generate aspartic acid residues (36). However, deamidation ofthe asparagine side chain through unregulated catalysis often has aT1/2 of at least 24 h. Considering that IBDV has an infection lifecycle of only 24 h, mature VP1 should retain the original asparagineresidue. Close inspection of the IBDV VP1 crystal structure revealsthat the side chain of Asn-403 is stacked against a Tyr (Tyr-405) sidechain, which is strictly conserved in all birnavirus polymerases (SIFigs. 6b and 7). The electron-rich aromatic ring may transfer someelectrons to the side-chain carbonyl of Asn-403, therefore enablingVP1 to bind not just Mn2�, but also the weak-chelating metal Mg2�.In other polymerases from dsRNA and �ssRNA viruses, theposition of Tyr-405 in VP1 is generally occupied by a hydrophobicresidue (Fig. 3). HIV-RT has a Tyr at this position, but its side chainis flipped away from the aspartate side chain in the enzyme–substrate complex (37).

The structure of our model VP1 elongation complex also shows

that Glu-421, the second acidic residue from motif A, superimposesonto Asp-590 in �3 and may be responsible for discriminating rNTPagainst dNTP (Fig. 3b) (30, 38). The side chain of glutamic acidresidue is one carbon bond longer than that of the aspartic acidresidue, which is strictly conserved in all other RdRps (Fig. 3a).Therefore, for Glu-421 to properly interact with the 2�-OH of rNTPin VP1, either the nucleotide substrate or the conserved acidresidue needs to adopt a slightly different spatial position comparedwith that in �3. Ser-484 from motif B may also play an importantrole in nucleotide selection, as discussed above. The rNTP mayaccess the active site through a positively charged tunnel in the backof the VP1 molecule. The positively charged substrate tunnel islined by a number of basic amino acid residues from the fingertip,palm, and thumb (Fig. 4). The modeled complex also reveals atemplate channel on top of the fingers, surrounded by residues fromfingers and thumb (Fig. 4). Many basic residues, such as K251, K255,R265, and K267, are located at the outer opening of the channel,forming a positively charged pore leading to the active site. Ile-337from motif F packs its hydrophobic side chain against the templatebase at the i � 1 position (Fig. 4). An Ile/Leu/Val is conserved inall RNA polymerases (Fig. 3a) and may be important for main-taining polymerase fidelity by ensuring proper base pairing betweentemplate and the nucleotide substrate (3, 37). The base stackingeffects of Ile-337 also force the template strand to make a �90° turnupon entering the active site. At the downstream end of the productexit channel, the sugar-phosphate backbone of nascent strand facesthe second �-helix in the thumb (T�2). This �-helix has been shownto interact with the minor groove of the dsRNA product duplex inseveral polymerases. The end of the VP1 product exit channel issurrounded by negatively charged residues, which may have theeffect of making translocation more efficient because of chargerepulsion.

In our structure, the product exit channel is partially blocked(Fig. 1a). At the beginning of the C-terminal domain, a 27-residueloop (from residues 658–684), which contains three short �-helices,protrudes into the polymerase active site from the downstream endof the active site canyon. This C-terminal protrusion, or theC-terminal plug, occupies the path of the template strand and wouldhave to be dislodged from its current location as the nascent RNAchain grows. Indeed, the molecular surface area buried by theC-terminal plug is only �900 Å, suggesting that it binds weakly tothe product exit channel and could be easily displaced. As shown inFig. 4, graphical deletion of the complete C-terminal domainreveals an unobstructed channel that fits nicely onto the dsRNAproduct duplex.

During transcription, the viral dsRNA duplex has to be separatedso that the negative-strand RNA can be used as a template for VP1.A predominantly positively charged groove is observed next to thetemplate entry channel (SI Fig. 8). This groove, which is formed atthe interface between the N-terminal domain and the polymerasethumb subdomain, may function to stabilize the non-template-strand RNA during the initiation of transcription. The non-template-strand RNA becomes the viral messenger, as birnavirustranscription occurs in a semiconservative fashion. A similar pos-itively charge groove is also noted for the polymerase from dsRNAphage �6 (7).

Functional Significance of the Active Site Asparagine. With so manyRdRp sequences currently available, it is clear that natural occur-rences of the Asp 3 Asn variation are extremely rare (15, 39).When the metal specificity was tested for RNA synthesis usingeither Mg2� or Mn2�, the wild-type VP1 was more active inpresence of Mn2� than Mg2� at 2 mM concentration (SI Fig. 5b,lanes 2 and 4). This result suggests that Mn2� is indeed morestrongly coordinated by the active site Asn in IBDV VP1. To furtherelucidate the functional significance of Asn-403, the active siteasparagine residue, we produced an IBDV VP1 mutant by replacingthe -401ADN- motif with the sequence -GDD-. As a result, the

7388 � www.pnas.org�cgi�doi�10.1073�pnas.0611599104 Pan et al.

mutant active site contains three aspartate residues, like otherdsRNA and ssRNA virus polymerases. When the 401GDD mutantwas subjected to in vitro polymerase assays in the presence of Mg2�,we found that it was able to replicate virus-specific RNA moreefficiently than the wild-type protein (SI Fig. 5b, lanes 7 and 9). Therelatively weak polymerase activity of the wild-type VP1 suggeststhat the initial occurrence of the Asp 3 Asn change at the 403position in birnavirus VP1 may enable birnaviruses to slow theirgrowth kinetics, therefore modulating the virulence and facilitatingvirus spread.

The Implication of VP1 Protein Priming. In the VP1 crystal structure,the putative guanylylation site residue S166 is present near thejunction of the N-terminal domain and the polymerase fingers. Theguanylylation site residue in IBDV VP1 has been tentativelymapped to S166 by sequence alignment, as the guanylylation site inthe homologous infectious pancreatic necrosis virus VP1 has beendetermined to be S163 using peptide digestion and site-directedmutagenesis (17). Surprisingly, S166 in IBDV VP1 is at least 23 Åfrom the polymerase active site in the same molecule. S166 alsoappears inaccessible for modification by the polymerase active siteof a different molecule because it faces the interior of the active sitecanyon. Therefore, the template-independent guanylylation ofIBDV VP1 may be catalyzed by a second active site and not by thepolymerase active site, a situation probably different from polio-virus. After guanylylation, significant conformational changes inthe N-terminal domain may occur such that Ser-166 and theassociated guanylyl group can bind to the polymerase active site forterminal initiation. Indeed, guanylylated VP1 proteins producedcrystals with a space group different from that of native VP1 (datanot shown), suggesting that some conformational changes may havetaken place.

The C-terminal domain of VP1 may function to prevent back-primed RNA synthesis during protein priming. The 27-residueC-terminal plug of VP1 occupies a location similar to the �-flap inHCV polymerase and the C-terminal domain of �6 polymerase (7,30). Rhinovirus polymerase, which initiates RNA synthesis viaprotein priming, also has a C-terminal peptide that protrudes intothe active site (9). Site-directed mutagenesis studies on �6 andHCV polymerases suggest that these elements may serve importantroles in RNA synthesis by preventing template back-priming topromote the synthesis of full-length genome (40). Indeed, only veryweak back-primed RNA synthesis activity is detected for isolatedVP1 in in vitro assays (41).

Evolutionary Relationship Between dsRNA Viruses and ssRNA Viruses.Birnaviruses are similar to other dsRNA viruses in that they possessa dsRNA genome and a T � 13 capsid structure (20). However, thegenome organization of birnaviruses shows several major differ-ences as compared with those of typical dsRNA viruses. Forexample, most dsRNA viruses contain a linear, segmented genomethat is either free or capped at the 5� end. Birnaviruses, however,have a VPg-linked genome, and initiate transcription/replication byprotein priming (SI Fig. 9). In addition, whereas most dsRNAviruses produce mRNA transcripts coding single protein products,birnaviruses encode a polyprotein in gene segment A (SI Fig. 9).Both the protein priming mechanism and the polyprotein codingstrategy are common features of �ssRNA viruses. The recentlyreported crystal structure of birnavirus capsid shows that the majorcapsid protein adopts a fold similar to those of noda- and tetravi-ruses, revealing an evolutionary link between birnaviruses and�ssRNA viruses (20). Here, by showing that IBDV VP1 adopts atopology that is also likely to be adopted by polymerases fromThosea asigna virus (TaV) and Euprosterna elaeasa virus (EeV)(15), both �ssRNA viruses in the alphavirus superfamily, ourresults lend further support for the evolutionary hypothesis. It isimportant to note, however, that we assume that horizontal genetransfer events have not occurred during birnavirus evolution.

Because TaV and EeV possess the conserved -GDD- sequence inmotif C, and birnavirus VP1 has the altered -ADN- sequence, it islikely that birnaviruses originated from a TaV/EeV-like ancestor inwhich the polymerase permutation had already taken place.

Except for birnaviruses, the permuted polymerase fold has notyet been found in other dsRNA viruses. Birnaviruses also differfrom other dsRNA viruses by lacking the ‘‘T � 2’’ icosahedralcapsid structure. Despite these differences, intact birnavirus parti-cles can readily produce mRNAs, an activity that is often assumedby the ‘‘T � 2’’ core in other dsRNA viruses. How do birnavirusesdeliver their messengers to the cytosol? Is birnavirus VP1 (VPg)removed from the 5� end of viral mRNAs after the mRNAs aresynthesized? What functional roles do the capsid proteins have inbirnavirus RNA synthesis? Further structural and biochemicalstudies of birnavirus RNA replication and transcription shouldprovide more insights into how birnaviruses are related to otherdsRNA viruses.

Materials and MethodsCloning, Protein Expression, and Purification. The IBDV (strain D78)VP1 coding region was cloned into pFastBacHTa (Invitrogen). Therecombinant baculovirus produces a VP1 protein with six histidineresidues followed by a TEV recognition site at the N terminus.IBDV VP1 protein was expressed by infecting Sf21 insect cells atthe multiplicity of infection (MOI) of 2. Infected cell pellet wasresuspended and sonicated in lysis buffer containing 50 mMTris-HCl pH7.5, 300 mM NaCl, 5 mM imidazole, 10% glycerol, 17�g/ml PMSF, 1 �g/ml leupeptin, 1 �g/ml pepstatin, and 5 mM�-mercaptoethanol, and the lysate was clarified by centrifugation at20,000 � g for 20 min. Recombinant VP1 was purified by chroma-tography using Ni-NTA affinity, a HiTrap Heparin–Sepharosecolumn, a Superdex-200 column, and a Mono Q column (Amer-sham). To obtain VP1c, full-length VP1 samples collected from theHeparin column were subjected to gentle chymotrypsin digestion at1:20 protease-to-protein molar ratio at 37°C for 10 min. Thereaction was stopped by adding PMSF to a final concentration of100 �g/ml before the next Superdex-200 chromatography step. Thepurified protein was at least 95% pure, as judged by SDS/PAGEstained with Coomassie blue.

N-terminal sequencing of VP1c indicated that a cleavage oc-curred after Tyr-18 at the N terminus. A mass spectrometryexperiment (ESI-MS) performed in parallel showed that the mo-lecular mass of the truncated fragment was 88.423 kDs, suggestingthat 68 amino acids had been removed from the C-terminal endafter Phe-810. The calculated molecular mass of VP1 polypeptidefrom residue 19 to 810 was 88.299 kDa. The �124-Da differencefalls within experimental error for proteins of this mass. Both massspectroscopy and N-terminal sequencing experiments were per-formed at the Protein Chemistry Core Laboratory at the BaylorCollege of Medicine (Houston, TX).

Crystallization and Data Collection. Crystals of VP1c in the spacegroup P6122 (Table 2) were obtained by the hanging-drop vapordiffusion method. The crystallization drop contained 2 �l ofrecombinant protein and 0.8 �l of well solution containing 100 mMHepes (pH 7.5) and 1.3 M sodium malonate. The best crystals hadthe shape of hexagonal plates and could grow to 0.4 � 0.4 � 0.2mm3 in size at 20°C in 3 days. The addition of 10 mM sodium iodideto the crystallization drop was critical, as it effectively reduced thenumber of nucleation sites. Each data set was collected from asingle frozen crystal that had been transferred in small incrementalsteps to the cryo solution (containing 50 mM Hepes, pH 7.5, 1.3 Msodium malonate, 30% glycerol, 5 mM DTT) before flash freezingin a boiling nitrogen stream at 100K. Heavy atom derivatives wereprepared by soaking crystals in cryo solution containing heavy atomcompounds at various concentrations (Table 2). All diffraction datawere collected by using the Rigaku RU-H3R X-Ray diffractionsystem equipped with the R-AXIS IV�� Image plates. Diffraction

Pan et al. PNAS � May 1, 2007 � vol. 104 � no. 18 � 7389

BIO

CHEM

ISTR

Y

data reduction and merging were performed by using DENZO andSCALEPACK (42).

Structure Determination. A combination of multiple isomorphousreplacement (MIR) and anomalous scattering (AS) from two heavyatom derivatives was used to solve the phase problem (Table 2). Thefigure of merit was 0.4509 for data from 30- to 2.5-Å resolution.Experimental phases were further improved by solvent flipping andhistogram matching using the program SHARP (43). The programO (44) was used for model building. CNS (45) was used forcrystallographic refinement with the idealized amino acid param-eters (46). A total of 492 water molecules were placed automaticallyin CNS. The final protein model had an Rfree � 25% and an Rwork �22%. According to the program PROCHECK (47), �87% of theresidues were in the most favored regions of a Ramachandran plot,and two were in disallowed regions. The final model contains 774amino acid residues out of a total of 878 present in VP1. Thoseabsent are the first 30 residues at the N terminus, of which the first

18 have been eliminated by proteolysis, and 74 residues from the Cterminus, of which the last 68 have been proteolytically removed.Moreover, an internal region from residues 603 to 611 is disordered.

Structure Superposition and RNA Modeling. Structure superpositionwas performed in O using the coordinates of the reovirus �3elongation complex (PDB ID code 1N35), �6 polymerase initiationcomplex (PDB ID code 1HHT), RHDV polymerase (PDB ID code1KHW), BVDV polymerase (PDB ID code 1S49), Rhinoviruspolymerase (PDB ID code 1TP7), HCV initiation complex (PDBID code 1GX5), Norwalk virus polymerase (PDB ID code 1SH3),and the HIV-1 RT catalytic complex (PDB ID code 1RTD). Thetransformation matrices were calculated by aligning 61 C� atomsfrom the five polymerase motifs (from A to E) in the palmsubdomain (Fig. 3).

In Vitro Self-Guanylylation and Polymerase Activity Assays. In vitroself-guanylylation assays were performed by using procedures sim-ilar to those described by Dobos (26). In short, pure protein samples(10 �g) were incubated with 50�Ci [�-32P]rGTP (3,000Ci/mmol)(0.02 mM) in 10 �l reaction buffer containing 50 mM Tris-HCl(pH7.5), 100 mM NaCl, 5% glycerol, 2 mM MgCl2, and 5 mM DTTat 37°C for 15 min. For replication assays, VP1 proteins (10 �g),rNTPs (0.5 mM ATP, CTP, UTP, and 0.01 mM GTP), RNasin (4units), RNA template (20 ng), and 10 �Ci [�-32P]GTP (3,000Ci/mmol) were mixed in 20 �l of reaction buffer [40 mM Tris, (pH8.0), 125 mM NaCl, 2 mM MgCl2 or MnCl2, 0.01 mM EGTA, 5 mMDTT, 0.01% Triton X-100], and incubated for 4 h at 37°C. Templatewas either the (�)RNA of segment A or a ssDNA from �x174.Reaction products were then mixed with SDS sample buffer,incubated at 70°C for 5 min, and loaded onto a 3–10% SDS/PAGEgel. Electrophoresis was carried out at 150 V until a prestained70-kDa band ran off the gel (�3 h).

Preparation of Figures. Secondary structure elements were assignedusing the program DSSP (48). Figures were prepared by usingMolscript (49), PyMOL (W. L. Delano, www.pymol.org) and Spock(J. A. Christopher, http://quorum.tamu.edu/spock).

We thank John Bruning, Douglas Mata, Max Nibert, Yousif Shamoo,and B. V. Venkataram Prasad for critical reading of the manuscript. Thiswork is supported by grants from the Welch Foundation (C-1565),National Science Foundation (EEC0118007), and National Institutes ofHealth (AI065733).

1. Tao Y, Farsetta DL, Nibert ML, Harrison SC (2002) Cell 111:733–745.2. Hansen JL, Long AM, Schultz SC (1997) Structure 5:1109–1122.3. Lesburg CA, Cable MB, Ferrari E, Hong Z, Mannarino AF, Weber PC (1999) Nat Struct Biol

6:937–943.4. Love RA, Maegley KA, Yu X, Ferre RA, Lingardo LK, Diehl W, Parge HE, Dragovich PS,

Fuhrman SA (2004) Structure (Cambridge) 12:1533–1544.5. Ng KK, Pendas-Franco N, Rojo J, Boga JA, Machin A, Alonso JM, Parra F (2004) J Biol Chem

279:16638–45.6. Ng KK, Cherney MM, Vazquez AL, Machin A, Alonso JM, Parra F, James MN (2002) J Biol

Chem 277:1381–1387.7. Butcher SJ, Grimes JM, Makeyev EV, Bamford DH, Stuart DI (2001) Nature 410:235–240.8. Choi KH, Groarke JM, Young DC, Kuhn RJ, Smith JL, Pevear DC, Rossmann MG (2004) Proc

Natl Acad Sci USA 101:4425–4430.9. Appleby TC, Luecke H, Shim JH, Wu JZ, Cheney IW, Zhong W, Vogeley L, Hong Z, Yao N

(2005) J Virol 79:277–288.10. Bressanelli S, Tomei L, Rey FA, De Francesco R (2002) J Virol 76:3482–3492.11. Duncan R, Mason CL, Nagy E, Leong JA, Dobos P (1991) Virology 181:541–552.12. Shwed PS, Dobos P, Cameron LA, Vakharia VN, Duncan R (2002) Virology 296:241–250.13. Poch O, Sauvaget I, Delarue M, Tordo N (1989) EMBO J 8:3867–3874.14. Koonin EV (1991) J Gen Virol 72:2197–2206.15. Gorbalenya AE, Pringle FM, Zeddam JL, Luke BT, Cameron CE, Kalmakoff J, Hanzlik TN,

Gordon KH, Ward VK (2002) J Mol Biol 324:47–62.16. Spies U, Muller H, Becht H (1987) Virus Res 8:127–140.17. Xu HT, Si WD, Dobos P (2004) Virology 322:199–210.18. Bottcher B, Kiselev NA, Stel’Mashchuk VY, Perevozchikova NA, Borisov AV, Crowther RA

(1997) J Virol 71:325–330.19. Caston JR, Martinez-Torrecuadrada JL, Maraver A, Lombardo E, Rodriguez JF, Casal JI,

Carrascosa JL (2001) J Virol 75:10815–10828.20. Coulibaly F, Chevalier C, Gutsche I, Pous J, Navaza J, Bressanelli S, Delmas B, Rey FA (2005)

Cell 120:761–772.21. Revet B, Delain E (1982) Virology 123:29–44.22. Spies U, Muller H (1990) J Gen Virol 71:977–981.

23. Calvert JG, Nagy E, Soler M, Dobos P (1991) J Gen Virol 72:2563–2567.24. Muller H, Nitschke R (1987) Virology 159:174–177.25. Salas M (1991) Annu Rev Biochem 60:39–71.26. Dobos P (1993) Virology 193:403–413.27. Magyar G, Chung HK, Dobos P (1998) Virology 245:142–150.28. Dobos P (1995) Virology 208:19–25.29. Bruenn JA (2003) Nucleic Acids Res 31:1821–1829.30. Bressanelli S, Tomei L, Roussel A, Incitti I, Vitale RL, Mathieu M, De Francesco R, Rey FA

(1999) Proc Natl Acad Sci USA 96:13034–13039.31. Joyce CM, Steitz TA (1994) Annu Rev Biochem 63:777–822.32. Vazquez AL, Alonso JM, Parra F (2000) J Virol 74:3888–3891.33. Lohmann V, Korner F, Herian U, Bartenschlager R (1997) J Virol 71:8416–8428.34. Jablonski SA, Morrow CD (1995) J Virol 69:1532–1539.35. Cohen J (1975) Biochem Biophys Res Commun 62:689–695.36. Wright HT (1991) Protein Eng 4:283–294.37. Huang H, Chopra R, Verdine GL, Harrison SC (1998) Science 282:1669–1675.38. Gao G, Orlova M, Georgiadis MM, Hendrickson WA, Goff SP (1997) Proc Natl Acad Sci USA

94:407–411.39. Bruenn JA (1991) Nucleic Acids Res 19:217–226.40. Laurila MR, Salgado PS, Stuart DI, Grimes JM, Bamford DH (2005) J Gen Virol 86:521–526.41. von Einem UI, Gorbalenya AE, Schirrmeier H, Behrens SE, Letzel T, Mundt E (2004) J Gen

Virol 85:2221–2229.42. Otwinowski Z, Minor W (1997) Methods Enzymol 276:307–326.43. Bricogne G, Vonrhein C, Flensburg C, Schiltz M, Paciorek W (2003) Acta Crystallogr D

59:2023–2030.44. Jones TA, Zou JY, Cowan SW, Kjeldgaard (1991) Acta Crystallogr A 47:110–119.45. Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS,

Kuszewski J, Nilges M, Pannu NS, et al. (1998) Acta Crystallogr D 54:905–921.46. Engh RA, Huber R (1991) Acta Crystallogr A 47:392–400.47. Laskowski RA, MacArthur MW, Moss DS, Thomton JM (1993) J Appl Crystallogr 26:283–291.48. Kabsch W, Sander C (1983) Biopolymers 22:2577–2637.49. Kraulis P (1991) J Appl Crystallogr 24:946–950.

Table 2. IBDV VP1 data collection and refinement statistics

Data set Native EMTS K2PtCl4

Data collectionUnit cell dimensions, Å a � b � 122.1 a � b � 122.1 a � b � 121.9of space group P6122 c � 359.0 c � 358.4 c � 358.3Resolution, Å 30–2.5 30–2.5 30–2.5

No. of reflections* 584,198 (55,871) 863,414 (55,794) 1,325,449 (55,717)Completeness, %† 96.4 (72.7) 97.0 (77.1) 99.6 (98.4)Rmerge, %† 6.5 (33.8) 8.6 (50.1) 7.8 (39.0)Riso, %�� (15.0–3.0Å) - 18.8 (24.0) 17.1 (20.0)Rano, %† - 4.6 (23.3) 4.2 (19.4)Rcullis

† (centric�acentric) - 0.85�0.93 (MIR) 0.89�0.95 (MIR)Phasing power‡

(centric�acentric)- 1.12�0.92 (MIR) 0.95�0.87 (MIR)

Figure of merit - 0.4509 (MIR-AS)Refinement statistics

Resolution range, Å 30–2.5Rwork, % 21.6Rfree, % 25.4RMS of bond lengths

and angles0.006 Å, 1.3°

*Numbers in parenthesis are the number of unique reflections.†Numbers in parenthesis are for the highest resolution bins.‡Phasing power � rms(FH�E), where E is the lack of closure error.

7390 � www.pnas.org�cgi�doi�10.1073�pnas.0611599104 Pan et al.