key: cord-0929526-rvgcalip
authors: Gilman, Morgan S.A.; Liu, Cheng; Fung, Amy; Behera, Ishani; Jordan, Paul; Rigaux, Peter; Ysebaert, Nina; Tcherniuk, Sergey; Sourimant, Julien; Eléouët, Jean-François; Sutto-Ortiz, Priscila; Decroly, Etienne; Roymans, Dirk; Jin, Zhinan; McLellan, Jason S.
title: Structure of the Respiratory Syncytial Virus Polymerase Complex
date: 2019-09-19
journal: Cell
DOI: 10.1016/j.cell.2019.08.014
sha: 94905c1dd5156e8390911f5fd65a93fe7f049518
doc_id: 929526
cord_uid: rvgcalip

Numerous interventions are in clinical development for respiratory syncytial virus (RSV) infection, including small molecules that target viral transcription and replication. These processes are catalyzed by a complex comprising the RNA-dependent RNA polymerase (L) and the tetrameric phosphoprotein (P). RSV P recruits multiple proteins to the polymerase complex and, with the exception of its oligomerization domain, is thought to be intrinsically disordered. Despite their critical roles in RSV transcription and replication, structures of L and P have remained elusive. Here, we describe the 3.2-Å cryo-EM structure of RSV L bound to tetrameric P. The structure reveals a striking tentacular arrangement of P, with each of the four monomers adopting a distinct conformation. The structure also rationalizes inhibitor escape mutants and mutations observed in live-attenuated vaccine candidates. These results provide a framework for determining the molecular underpinnings of RSV replication and transcription and should facilitate the design of effective RSV inhibitors.

Respiratory syncytial virus (RSV) remains a leading cause of bronchiolitis and hospitalization, especially of infants. Gilman et al. present a 3.2-Å cryo-EM structure of the RSV L polymerase in complex with the P phosphoproteincomponents of the core viral replication machinery that represent an attractive target for the development of therapeutic agents.

Respiratory syncytial virus (RSV) is a ubiquitous viral pathogen that infects nearly all children before 2 years of age (Glezen et al., 1986) . It is estimated that RSV causes more than 60,000 in-hospital deaths per year in children less than 5 years old, with the majority of deaths occurring in developing regions (Shi et al., 2017) . In addition, RSV infection can lead to serious complications in the elderly and high-risk adults and causes a disease burden similar to non-pandemic influenza A in these populations (Falsey et al., 2005) . Currently, the only clinical intervention for RSV is passive prophylaxis with the monoclonal antibody Synagis, which reduces RSV-related hospitalization (The IMpact-RSV Study Group, 1998) . However, the high cost and moderate efficacy of Synagis restrict its use to high-risk infants in developed regions of the world (Homaira et al., 2014; Shahabi et al., 2018; The IMpact-RSV Study Group, 1998) . Thus, there is great interest in developing more effective therapeutic agents for RSV, including small-molecule antivirals.

RSV is an enveloped, negative-sense, single-stranded RNA virus that belongs to the Pneumoviridae family (Afonso et al., 2016) . Only three viral proteins are absolutely required for replication of the RSV genome: the nucleoprotein (N), the large protein (L), and the phosphoprotein (P) (Grosfeld et al., 1995; Yu et al., 1995) . N associates with viral RNA to form a tightly woven helical assembly that protects the RNA from cellular nucleases and recognition by the innate immune system (Bakker et al., 2013; Tawar et al., 2009) . L harbors three conserved enzymatic domains: the RNA-dependent RNA polymerase (RdRp) domain, the polyribonucleotidyl transferase (PRNTase or capping) domain, and the methyltransferase (MTase) domain, which catalyzes cap methylation. These activities are all potential targets for inhibitor development. Nucleoside analogs that terminate RNA chain synthesis have been identified (Clarke et al., 2015; Deval et al., 2015; Wang et al., 2015) , and one such compound, ALS-8176, has shown efficacy in RSV-infected adults (DeVincenzo et al., 2015) . Several non-nucleoside small-molecule inhibitors have also been identified (Cockerill et al., 2019; Fearns and Deval, 2016) , and although some (for example, BI-compound D) are known to disrupt RNA cap addition (Liuzzi et al., 2005) , there are additional inhibitor classes for which the mechanism of action is not well understood (Duvall et al., 2016; McCutcheon et al., 2015) . P serves as an essential polymerase cofactor that tethers L to the nucleoprotein-RNA complex (García et al., 1993; Grosfeld et al., 1995; Yu et al., 1995) . P also acts as a chaperone that prevents the association of nascent N (N 0 ) with host cell RNAs Pereira et al., 2017; Tran et al., 2007) and is responsible for recruiting the M2-1 protein, a processivity factor that is required for efficient transcription of viral RNA (Blondot et al., 2012; Collins et al., 1996; Mason et al., 2003) . In addition, P recruits the cellular phosphatase PP1 to inclusion bodies to regulate viral transcription (Richard et al., 2018) . Thus, P plays critical roles in regulating RNA replication and transcription through its interactions with multiple proteins. Structurally, RSV P contains a central oligomerization domain that is predicted to form a tetrameric coiled coil (Castagné et al., 2004; Llorente et al., 2008) . Regions N-and C-terminal to the oligomerization domain are predicted to be intrinsically disordered and may only adopt defined conformations when bound to other proteins. The dynamic nature of RSV P has prevented determination of its structure (Pereira et al., 2017; Simabuco et al., 2011) ; thus, the molecular mechanisms by which P coordinates the activities of diverse viral components are not well understood.

To gain atomic-level information regarding RSV transcription and replication, we initiated structural studies of a purified polymerase complex comprising L and P. The resulting 3.2-Å cryoelectron microscopy (cryo-EM) structure reveals that P displays unique structural plasticity, with each monomer adopting a different conformation as it interacts with distinct regions on L. The variability in secondary structure of each P monomer indicates that this viral phosphoprotein exhibits characteristics of a ''transformer'' protein (Knauer et al., 2012) . In addition, the surface on the RSV L RdRp domain that is recognized by RSV P is similar to the region bound by the ribosomal S1 protein cofactor on the RdRp of the distantly related Qb bacteriophage polymerase, indicating that this interaction may be evolutionarily conserved. Our results also provide the first structural description of the RdRp and capping domains of RSV L and provide insights into the mechanism of viral escape from nucleoside analog and non-nucleoside inhibitors. Collectively, these results inform our understanding of viral polymerase complexes and should aid in the development of next-generation inhibitors of RSV infection.

RSV L Co-expressed with RSV P Is Biochemically Active Human RSV L was co-expressed with RSV P in Sf9 cells. The purified protein was analyzed by size exclusion chromatog-raphy, and SDS-PAGE analysis confirmed the presence of both L ($250 kDa) and P ($40 kDa) in the final sample (Figure 1A) . Densitometric analysis and known oligomeric states of the proteins suggested a stoichiometry of 1:4 (L:P). To verify that the L-P complex was functional, RdRp activity was measured in an RNA template-dependent primer-extension assay. In the presence of GTP, ATP, and a 33 P-CTP, L-P incorporated 67-fold more a 33 P-CTP into the growing primer than what was detected in the control reaction ( Figure 1B ). L-P activity was also measured in a nucleotide-incorporation assay. When incubated with 33 P-GTP, the polymerase incorporated one 33 P-GMP (+1) into the primer ( Figure 1C , lane a). With the addition of 33 P-GTP and ATP, the enzyme extended the primer by three nucleotides (+3), GAA, complementary to the RNA template ( Figure 1C , lane b). When incubated with 33 P-GTP, ATP, and CTP, the enzyme extended the primer by seven nucleotides (+7), GAACAAA ( Figure 1C , lane c). As a control, RSV L-P added only four nucleotides (+4) when incubated with 33 P-GTP, ATP, and ALS-8112-TP, the active form of a known RSV polymerase nucleoside inhibitor that causes chain termination ( Figure 1C , lane d) . The nonprocessive, abortive pattern of nucleotide addition and the chain termination property of ALS-8112-TP observed with RSV L-P was consistent with results reported previously , indicating that the recombinant RSV L-P has RNAdependent RNA polymerase activity.

Because the L protein harbors multiple enzymatic domains, we sought to further validate the recombinant RSV L-P protein by measuring its MTase activity. The transfer of tritiated methyl groups from S-adenosyl methionine (SAM) to synthetic RNA molecules derived from the 5 0 end of RSV mRNA was measured using a filter-binding assay. The L-P protein exhibited MTase activity for three synthetic GpppG RNAs of different lengths, with activity 7-to 10-fold higher than a control reaction lacking the synthetic RNAs ( Figure 1D ). Collectively, these results indicated that the recombinant L-P complex was functionally active and suitable for structural analysis by cryo-EM. Molecular weights (in kilodaltons) of the ladder are shown on the left, and the L and P bands are labeled on the right. (B) RNA synthesis activity of the RSV L-P complex was tested in a primer extension assay using an 11-mer RNA template and biotinylated 4-mer RNA primer. Primers were captured on streptavidin-coated plates, and extension was measured by incorporation of a 33 P-CTP. Bars indicate the mean, and error bars depict the SD of one independent experiment with 16 technical replicates (*p < 0.001, Mann-Whitney U test). (C) Polyacrylamide sequencing gel showing the results of the nucleotide incorporation assay for a single nucleotide (lane a) and multiple nucleotides (lanes b-d). ALS-8112-TP is a CTP analog that terminates elongation. (D) Transfer of tritiated methyl groups from S-adenosyl methionine (SAM) molecules to a set of synthetic RNAs of varying length (5-, 9-, and 15-mers) that mimic the 5 0 end of RSV mRNA was measured using a filter-binding assay. Bars indicate the mean, and error bars depict the SD of four independent experiments. Asterisks indicate means that were statistically different from that of the control sample as measured by Dunnett's test (p < 0.001).

Determination of the 3.2-Å Cryo-EM Structure of RSV L in Complex with Tetrameric P The L-P complex was plunge-frozen on UltrAuFoil 1.2/1.3 grids (Russo and Passmore, 2014) , and a dataset was collected using a Titan Krios equipped with a K2 Summit detector ( Figure S1 ; Table 1 ). After motion correction, contrast transfer function (CTF) estimation, 2D classification, and iterative rounds of heterogeneous 3D refinement, a total of 241,669 particles remained. Homogeneous 3D refinement of this particle stack resulted in a 3.4-Å map. After initial building of the L model and docking of the closely related human metapneumovirus (hMPV) P oligomerization domain structure into the map (Leyrat et al., 2013) , it was possible to identify the portions of the density that corresponded to the P monomers. We then generated a map in which these regions were deleted. A low-pass filter was applied to this new map as well as to the original map, and both were used as input volumes for a final round of heterogenous 3D refinement to remove particles in which the tetrameric P had dissociated. The resulting stack of 196,720 RSV L-P particles was subjected to a homogeneous, non-uniform 3D refinement that yielded the final 3.2-Å map. The low sequence identity of vesicular stomatitis virus (VSV) L prevented use of the VSV L structure as a starting model, but this structure was used to guide manual building of RSV L. Although the sharpened map was used for the majority of model building, the C-terminal portions of two P monomers were difficult to place in this map because of over-sharpening.

However, these regions were visible in the unsharpened maps; therefore, both maps were used in combination to build these regions of the polymerase complex.

A total of 1,361 residues of L (residues Asn11-Leu133, Thr184-Gly618, Arg627-Ser658, and Asn690-Glu1460) were built into the 3.2-Å map ( Figure 2 ). Based on homology with the VSV L structure, residues 1-968 belong to the RdRp domain, and residues 969-1460 belong to the PRNTase (capping) domain. Although the MTase domain, connector domain (CD), and C-terminal domain (CTD) of our construct were present when analyzed by SDS-PAGE after size exclusion chromatography, and the protein exhibited MTase activity in vitro, these domains were not visible in our map ( Figure S2 ). This finding is not unexpected, however, given that the related VSV L is known to sample multiple states in which these three domains display various degrees of association with the RdRp and PRNTase domains (Rahmeh et al., 2010) . In addition to the L core, we could build portions of four discrete monomers of P, ranging in length from 53-98 residues. The longest monomer of P spans residues 131-228, corresponding to the oligomerization domain and most of the C-terminal region of P.

The RSV P Tetramer Grips L Using an Unusual Tentacular Arrangement As expected from previous experiments, residues 131-151 of RSV P formed a tetrameric coiled coil (Castagné et al., 2004; Llorente et al., 2006; Pereira et al., 2017) . Historically, regions outside of this oligomerization domain have not been structurally well characterized because of their intrinsic flexibility (Longhi et al., 2017) . However, the interaction of P with L stabilized ordered conformations of the C-terminal regions of P, allowing them to be resolved for the first time. Perhaps the most striking feature of the tetrameric P structure is the large degree of variation in the conformation of each of the four P monomers ( Figure 3 ) (P 1 , P 2 , P 3 , and P 4 ). These different conformations allow P to wrap around the RdRp of L in a tentacular fashion ( Figure 4 ). Although P 1 and P 4 make extensive contacts with L, P 3 makes minimal contacts with L, and P 2 interacts almost exclusively with the other P monomers. The unique organization adopted in this complex allows a relatively small region of P (less than 100 residues) to contact more than 11 regions of the primary L sequence and bury more than 4,000 Å 2 of surface area on L ( Figures 4A and 4B ).

The interaction of P with L is driven by a number of hydrophobic and electrostatic interactions ( Figure 4B ). The majority of electrostatic interactions are made between residues 167-179 of P 1 and the RdRp domain of L. Two positively charged P 1 residues, Arg167 and Arg174, form salt bridges with L residues Asp718 and Glu722 ( Figure 4C ). In addition, two negatively charged residues on P 1 , Glu176 and Glu179, form salt bridges with Arg523 and Lys529 on L. We predict that mutation of these residues would disrupt the L-P interaction and decrease the activity of L. Consistent with this hypothesis, two of these charged P residues (Arg174 and Glu176) have been shown previously to be critical for the activity of L in a minigenome assay (Lu et al., 2002) . In addition, a temperature-sensitive virus with an E176G substitution in P reverted the glycine to a negatively charged residue when the virus was passaged at physiological temperatures (Lu et al., 2002) . The RSV L-P structure thus provides a molecular explanation for the decreased activity of L because of these substitutions in P. The P 4 monomer, for which the largest number of residues could be built, contacts a completely different set of L residues than P 1 . Two residues on P 4 (Ile181 and Leu198) form a hydrophobic cap above Phe452 on the b6 strand of L ( Figure 4D ). Additionally, a region of P 4 forms a b-hairpin that hydrogenbonds with b6 of L to form a small b sheet ( Figure 4D ). In P 1 and P 2 , these same residues are within an a helix, and in P 3 , they are disordered, indicating that the conformation of this region of P is particularly malleable. Furthermore, at the base of the loop connecting the b strands of the P 4 hairpin, Arg163 forms a salt bridge with Glu765 of L, and the main chain of Gly165 forms a hydrogen bond with Tyr772 of L ( Figure 4D ). The C-terminal region of P 4 makes additional contacts with L, including hydrogen bonds between Asn217 of P and Asn362 and Thr365 of L, as well as hydrophobic interactions between Leu223 of P and surrounding hydrophobic residues on L ( Figure 4E ). Leu223, along with Leu216 and Leu227, are three hydrophobic residues on the C terminus of P shown previously to contribute to L binding. Mutation of these residues disrupts polymerase activity , demonstrating that these interactions between L and P, which are critical for function of the polymerase, can be explained by the structure presented here.

In addition to the contacts made between P and L, extensive interactions between P monomers are also present. Contacts within the oligomerization domain were expected (Castagné et al., 2004; Llorente et al., 2006; Pereira et al., 2017) , but the structure reveals that interactions between P monomers also occur outside of the oligomerization domain. The C-terminal a helices of P 2 and P 3 rest on top of P 1 to form a three-helix bundle ( Figure 4A ). In this arrangement, Ala169, Ile173, and Ile178 of P 1 are capped by hydrophobic residues on P 2 and P 3 . To validate the importance of these residues as well as of those described above, we generated a panel of 37 variants of P in which one or two residues were mutated to alanine and tested these variants in a minigenome replication assay. Variants with substitutions of residues between position 131 and 189 as well as position 205 showed less than 20% of the activity observed for wild-type P ( Figure S3 ). For the P variants with reduced activity, 16 of the substitutions mapped to regions involved in formation of the coiled coil in the oligomerization domain, 8 were in regions involved in the interaction between P and L, and 8 mapped to residues involved in the interactions between P 1 , P 2 , and P 3 that lie outside of the oligomerization domain. Collectively, these results demonstrate that the multifaceted interactions of P identified here are critical for its function in viral replication.

The RNA-dependent RNA polymerases of negative-sense RNA viruses contain four subdomains-the palm, fingers, thumb, and a subdomain thought to serve as a structural support (Figure 5A )-and six conserved sequence motifs (motifs A-F) (Já come et al., 2015; Poch et al., 1989) . The majority of these motifs are found in the palm subdomain ( Figure 5B ), which is composed of two a helices (a29 and a30) and a b sheet comprising 5 strands (b9, b12, b13, b14, and b15). A conserved glycine residue (Gly877) just after the motif E b sheet connects the palm and thumb subdomains and may allow the palm domain to exhibit small degrees of motion relative to the thumb, as demonstrated for the enterovirus polymerase (Gong and Peersen, 2010; Figure 5B ). Motif B forms one of two connections that are present between the palm and fingers. This motif is composed of a flexible loop that extends from the finger subdomain into the a29 helix of the palm. The presence of a glycine just before the a29 helix is known to be critical for catalysis in the RdRps of hepatitis C and encephalomyocarditis viruses (Lohmann et al., 1997; Sankar and Porter, 1992) . RSV harbors two additional glycine residues that form a GGxxG motif also found in the RdRp domains of L proteins from hMPV, Ebola, VSV, and other non-segmented negative-sense RNA viruses ( Figure S4 ). Thus, the palm subdomain is framed on either side by glycine residues that likely aid in flexibility of this region to accommodate the template RNA and incoming nucleoside triphosphates (NTPs) in the catalytic site (Garriga et al., 2013) .

The conserved catalytic motif (Gly810-Asp811-Asn812 [GDN]) of the RdRp is situated on the loop that connects b12 with b13 in motif C ( Figure 5C ). Another catalytic residue, Asp700, is located in motif A on the b9 strand. Together, Asp811, Asn812, and Asp700 are responsible for coordinating the two magnesium ions required for catalysis of phosphodiester bond formation (Noton et al., 2014) . In our map, no density for magnesium ions is observed, but this may be due to the resolution or lack of magnesium during purification. A conserved aromatic residue, Phe704, is located in motif A on the a26 helix ( Figure 5C ). The side chain of Phe704 is oriented away from the active site and toward the a helix of Motif B, where it is neighbored by three hydrophobic residues (Leu701, Phe708, and Trp785). The presence of an aromatic residue at position 704 is well conserved, but it was previously unclear whether this residue is required for nucleotide binding in the active site or for stability of the L protein (Já come et al., 2015) . The RSV L-P structure suggests the latter.

The majority of the RSV RdRp finger subdomain is a-helical, but a small b sheet composed of b7, b8, and b10 contains the final motif required for polymerase activity (motif F) ( Figure 5B ). Although a portion of the density in this region is disordered in our map and prevented us from building residues 609-626, the base of the b7-b8 hairpin and the Arg627 side chain are visible and positioned to interact with negatively charged phosphate groups in the active site ( Figure 5C ). In addition to its role in coordinating the template strand and the incoming NTP for catalysis, the finger subdomain also interacts with the structural subdomain to form an electropositive pore that likely serves as the entry point for incoming NTPs ( Figure S5) , similar to what has been described for other viral polymerases (Choi et al., 2004; Lu et al., 2008; Tao et al., 2002) .

Occurs by an Allosteric Mechanism Viral escape from the nucleoside inhibitor ALS-8176 has been mapped to the so-called ''Quad'' substitutions in Figure 3 . The RSV P Tetramer Wraps around L in a Tentacular Fashion RSV L is shown as a white molecular surface, and P monomers are shown as ribbons. The P 4 monomer is colored as a rainbow (blue to red, N to C terminus, respectively). All other monomers are colored to match P 4 . See also Figures S6 and S7. RSV L (M628L/A789V/L795I/I796V) . The RSV L-P structure reveals that three of these residues are buried on the a29 helix of Motif B, oriented away from the active site of the polymerase, with one of the residues (A789V) in close proximity to Phe704 of motif A ( Figure 5D ). Previous reports have suggested that the Quad substitutions act by increasing the ability of the RdRp to discriminate the altered structure of the cytidine analog from that of cytidine, preventing its incorporation into the growing RNA . Our structure suggests that the A789V, L795I, and I796V substitutions act allosterically by altering the conformation of the polymerase active site without directly contacting either native NTPs or the cytidine analog. This is consistent with the finding that motif B is involved in sensing the correct NTP and allosterically regulating the activity of the RdRp (Garriga et al., 2013) .

The PRNTase domain of RSV L is responsible for the addition of a 5 0 cap to nascent viral mRNAs (Braun et al., 2017) . In addition, the PRNTase domain contains the priming loop (1267-1282), which is thought to facilitate de novo initiation . In the structure determined here, the priming loop is flipped upward away from the RdRp domain ( Figures 6A and 6B) , removed from the RdRp active site. This is in contrast to what was observed previously in the VSV L structure, in which the priming loop reached out of the PRNTase domain toward the RdRp active site ( Figure 6B ; Liang et al., 2015) . For the VSV L structure, the authors suggested that the conformation observed represents the initiation state and that the priming loop as well as the CD, MTase, and CTD domains would need to reorient for elongation to occur. The alternate positioning of the priming loop in the RSV L structure, combined with the lack of observable density for the CD, MTase, and CTD domains, suggests that the structure determined here may resemble an elongation conformation of RSV L.

Sequence analysis of the L proteins of non-segmented, negative-sense RNA viruses has identified several conserved residues in the PRNTase domain that have also been assigned motifs (motifs A-E) (Li et al., 2008; Neubauer et al., 2016) . In RSV L, these residues are clustered around the center of the cavity formed between the RdRp and PRNTase domains ( Figure 6C ).

Motif D (also referred to as the His-Arg or HR motif) consists of two catalytic residues that are critical for cap formation: His1338 and Arg1339 (Braun et al., 2017) . The position of the priming loop in the RSV L structure results in a distance between motif D and motif B (also called the GxxT motif) of less than 5 Å ( Figure 6D ). Thus, flipping of the priming loop into a position likely associated with elongation is coupled with closing of the active site of the PRNTase domain into a compact conformation. Gly1264 marks the point around which the priming loop hinges (Figure 6B) , and a variant with a substitution to a less flexible amino acid (alanine) at this position has been shown previously to have defects in initiation of replication, elongation of viral RNA, and cap addition (Braun et al., 2017) . Collectively, these data suggest that motion of the priming loop is critical for all three processes, consistent with biochemical data that has demonstrated crosstalk between RNA synthesis and capping (Braun et al., 2017; Neubauer et al., 2016) .

In addition to Gly1264, Trp1262 may also play a role in positioning the priming loop ( Figure 6D ). The presence of an aromatic residue at this position is conserved in L proteins from multiple non-segmented, negative-sense RNA viruses ( Figure S4 ), likely because of hydrophobic interactions with residues N-terminal to motif E. However, the precise identity of this residue is not conserved, and the tolerance for different hydrophobic residues at this position depends on the viral background (Neubauer et al., 2016) . The RSV L structure reveals that a hydrogen bond between the indole group of Trp1262 and the hydroxyl group of Ser1390 may help to anchor motif B and motif D together to form the PRNTase active site. In VSV L, the Trp-Ser interaction is replaced with a hydrogen bond that is formed between Tyr1152 and Gln1270. Thus, although Trp1262 is not conserved across viruses, the interaction between these two regions is structurally preserved.

The non-nucleoside small-molecule inhibitor BI-compound D (BI-D) disrupts RNA cap addition, but its binding site on RSV L is currently unknown. Viruses that escape inhibition by BI-D harbor mutations encoding one of three substitutions in residues surrounding the PRNTase active site (I1381S, L1421F, or E1269D) (Liuzzi et al., 2005) . Ile1381 and Leu1421 are in close proximity to one another on the surface of L. Ile1381 contributes to a small pocket on the surface of the PRNTase domain that includes motifs B, D, and E ( Figure 6E ). Although Leu1421 does not form a part of the cavity wall, it contacts Ile1381, and a substitution to the larger phenylalanine side chain may indirectly alter the shape of this pocket through its interaction with Ile1381. The location of these residues near a surface-exposed pocket within the active site of the PRNTase domain suggests that this pocket could be the binding site of BI-D.

The substitution present in the third escape variant occurs at Glu1269, which is approximately 20 Å away from Ile1381 and Leu1421. Glu1269 forms hydrogen bonds with three residues on L (Thr1341, Lys1294, and Trp1297), anchoring the priming loop to motif C ( Figure 6F ). In this conformation, Glu1269 is oriented away from the PRNTase active site. Substitution of Glu1269 with the shorter aspartate residue present in the escape mutant would disrupt this hydrogen bonding network and likely change the conformation of the priming loop. This change would be expected to affect the organization of the proposed BI-D binding site because Motif B contributes to the surface pocket and is located on the priming loop, just two residues from Glu1269. Thus, it seems unlikely that Glu1269 directly contacts BI-D, and the structure suggests that the E1269D substitution results in allosteric changes to the BI-D binding site.

A number of live-attenuated RSV vaccine candidates have mutations in L that render the virus temperature sensitive. One virus, called cpts530/1030, has been used as a background strain for engineering viruses with high levels of attenuation and genetic stability (Whitehead et al., 1999) . The ''1030'' mutation in this virus refers to a substitution (Y1321N) that is located on a50 in the PRNTase domain, between motif C and the catalytic motif D. In the RSV L structure, Tyr1321 forms a hydrogen bond with (B) . Select residues are shown as sticks, with oxygen atoms colored red, nitrogen atoms colored blue, and sulfur atoms colored yellow. The Ca atoms of conserved glycine residues are shown as spheres. The panel on the right shows the same view, but with a UTP molecule modeled based on alignment of the palm subdomain active site with the hepatitis C virus RNA polymerase structure (PDB: 1NB6). (D) The same region as in (C), rotated by 180 to show three residues on motif B that are mutated in viruses that escape inhibition by ALS-8176. See also Figure S5 .

Lys1305 on a49. This bond may help lock the position of these two a helices relative to one another but likely could not form with the shorter side chain of an asparagine residue ( Figure 6G ). The catalytic residues of motif D, as well as two aromatic residues (Tyr1331 and Tyr1336) that stack on either side of Trp1297 of motif C, are both located on the loop just C-terminal to a50. Motion of a50 away from a49 would be expected to disrupt the interaction of Tyr1331 and Tyr1336 with Trp1297 and alter the conformation of the loop that contains the catalytic residues. It is likely that the temperature sensitivity of the 1030 mutant virus is due to destabilization of this region, which can be compensated by growth at reduced temperatures. Mutational analysis has also demonstrated that, although the Y1321N substitution moderately attenuates viral replication in a mouse model of infection, insertion of a negatively charged residue into this position (Y1321E) results in more dramatic attenuation (Luongo et al., 2012) . There are several negatively charged residues present on a48 and a49 that would repel a negatively charged residue if it were present on a50; thus, this substitution would be expected to have a more dramatic effect on the relative orientation of these three a helices.

Replication and transcription of negative-sense RNA viruses rely on specialized RNA-dependent RNA polymerases that must be carried within the virus. These RdRps are essential for viral survival and are structurally dissimilar to host-cell polymerases, making them excellent targets for antivirals. To efficiently replicate and transcribe viral RNA, RdRps typically require one or more cofactors. Although understanding the interactions between viral polymerases and their cofactors can provide additional targets for inhibitor development, structural characterization of these interactions has often proven difficult. The structure presented here provides the first structural characterization of a viral phosphoprotein in complex with a viral L protein.

The phosphoprotein of viruses in the Mononegavirales order are generally composed of an N-terminal domain, a central oligomerization domain, and a C-terminal domain and are known to tether multiple viral components to a centralized area where viral replication occurs (Cox and Plemper, 2015) . Interactions with viral proteins have primarily been mapped to the N-and C-terminal domains, which are thought to be intrinsically disordered, but (G) View of the region surrounding the a helix that is mutated in temperature-sensitive viruses. In all panels, select residues are shown as sticks, with oxygen and nitrogen atoms colored red and blue, respectively, and hydrogen bonds depicted as dashed lines. are capable of transiently adopting a-helical arrangements in solution or when bound to other proteins (Pereira et al., 2017; Selvaraj et al., 2018) . Previous work has indicated that this structural plasticity is required for RSV P function (Noval et al., 2016) . Intrinsically disordered proteins are thought to adopt ''fuzzy'' structures in which flexibility is retained even when these proteins interact with other binding partners (Arbesú et al., 2018) . The regions of P that are not contacting L are not visible in our cryo-EM map, indicating that unliganded P is indeed dynamic. However, we show that regions of P can become ordered when bound to L and that the same residues of P monomers can adopt distinct conformations, indicating that RSV P shares properties with the ''transformer'' proteins (Knauer et al., 2012) . This class of protein has also been identified in the Ebola virus, which uses different oligomeric states of the VP40 protein to assemble the viral matrix layer and to regulate viral transcription (Bornholdt et al., 2013) . The ability to utilize one protein for multiple interactions via structural transformations helps maximize the efficiency of viruses with limited genome sizes, and our results suggest that this might be a common phenomenon among RNA viruses.

For VSV L, an N-terminal fragment of the P was sufficient to favor the formation of a compact structure in which the CD, MTase, and CTD domains packed against the core of L Rahmeh et al., 2012) . However, VSV P was not clearly visible in the cryo-EM map obtained for the VSV L-P complex. In contrast, the RSV L-P structure described here shows that the oligomerization domain and C-terminal region of P, rather than the N-terminal region, interact with L. The P sequences of these two viruses vary substantially, and the region of VSV P used to stabilize the VSV L structure has low sequence homology to RSV P ( Figure S6 ). These results suggest that the mechanisms by which P influences the structural organization of L may differ between RSV and VSV.

The results of our structural analysis allow us to update the interaction map that has been generated for RSV P (Pereira et al., 2017) . Although structural studies have defined the interaction of M2-1 (Selvaraj et al., 2018) and N 0 (Esneau et al., 2019; Galloux et al., 2015) with N-terminal peptides of RSV P, interactions in the oligomerization domain and C-terminal region are less well understood. Biochemical data have implicated P residues 216-239 in binding to L ; however, the RSV L-P structure and minigenome assay results presented here demonstrate that, although these residues are important for binding to L, contacts between L and P are more extensive than previously thought and include the oligomerization domain and surrounding regions, particularly the charged residues from 167-179. It has also been suggested that the nucleocapsid and L binding sites on P may overlap (Lu et al., 2002; Pereira et al., 2017; Sourimant et al., 2015) . Because the four different P monomers adopt different conformations and make unique contacts with L, it is possible that P could be bound simultaneously by L and nucleocapsid, with different monomers mediating the distinct interactions with each. Determining whether simultaneous binding of L and nucleocapsid occurs by such a mechanism will likely require structural information for the entire replication complex.

Although the sequence of the phosphoproteins within the Mononegavirales order show a large degree of diversity, their function is well conserved. The phosphoproteins, as well as polymerase cofactors from other viral families, serve multiple functions, all of which rely on their ability to interact with the viral polymerase to recruit it to the site of viral replication. Comparison of our structure with that of the recently described severe acute respiratory syndrome coronavirus (SARS-CoV) polymerase complex reveals that the region on the RdRp domain of RSV L that is bound by three of the four P monomers is similar to the region on the SARS-CoV polymerase that is occupied by residues in an N-terminal extension of the polymerase ( Figure S7 ; Kirchdoerfer and Ward, 2019) . Perhaps even more strikingly, comparison of the RSV L-P complex structure with the structure of the replicase holoenzyme complex from bacteriophage Qb also reveals conserved features. During infection in bacteria, Qb recruits the host-cell ribosomal S1 protein, which is composed of six oligonucleotide-oligosaccharide-binding (OB) domains, to initiate viral replication (Wahba et al., 1974) . Two of these domains, OB1 and OB2, contact the Qb replicase core but use different motifs to contact discrete regions and wrap around the replicase core (Gytz et al., 2015; Takeshita et al., 2014) . Thus, although the structure of S1 and P share no homology, the modular domains of S1 and three of the monomers of P contact similar regions on the finger subdomain of their respective RdRp using a similar strategy ( Figure S7 ). Contacts made with this region on the RdRp domains of these three viral polymerases may simply act to tether the replication complex and increase the interaction with viral RNA. However, the conserved location of this interface implicates these cofactors in regulating or maintaining the conformation of the replication machinery. This role can be filled by a domain of the polymerase itself (coronavirus), a secondary viral protein (RSV), or even a hijacked host cell protein (Qb). Future studies will seek to determine how P influences the conformation of RSV L and further clarify the role of this evolutionarily conserved protein-binding interface in viral replication.

In addition to the interactions between P and L, our results provide the first structural information for the RdRp and capping domains of RSV L. The structure of the RdRp domain of RSV L is similar to that observed for VSV L , whereas the PRNTase domains of the two L proteins were observed in different states. In the previously determined VSV L structure, the priming loop was inserted into the RdRp domain, poised to serve a role in initiation. However, in the RSV L structure determined here, the priming loop is flipped away from the RdRp domain of RSV L, suggesting that the conformation identified here represents a non-initiation state and could more closely resemble the elongation conformation of the polymerase. The RSV L structure also rationalizes inhibitor escape mutants and mutations observed in live-attenuated vaccine candidates (Duvall et al., 2016; Liuzzi et al., 2005; Luongo et al., 2012; Whitehead et al., 1999) . Several of the substitutions that enable escape from nucleoside and non-nucleoside inhibitors appear to allosterically alter the conformation of the inhibitor binding sites, although additional structural studies of inhibitor-bound complexes are needed to fully understand these mechanisms of escape. Atomic-level knowledge of these mechanisms should facilitate the design of antivirals with increased barriers to escape. The RSV L-P structure also provides detailed information for protein-protein interactions that could serve as novel targets for inhibitor development. Thus, our results lay the foundation for the design of novel antivirals and for the improvement of existing small-molecule inhibitors for RSV.

Detailed methods are provided in the online version of this paper and include the following: 

Further information and requests for materials should be directed to and will be fulfilled by lead contact Jason S. McLellan (jmclellan@ austin.utexas.edu).

Protein production and purification Human RSV L protein (strain A2) with an N-terminal dual StrepTag and RSV P protein (strain A2) with a C-terminal 6x His-tag were coexpressed in Sf9 insect cells using pFASTbac Dual transfer vector (Life Technologies). The Sf9 cells expressing the L-P polymerase complex were lysed by two passes through a microfluidizer in Strep Buffer A (50 mM Tris-HCl, 300 mM NaCl, 10% glycerol, and 1 mM Tris (2-carboxyethyl) phosphine (TCEP), pH 8.0) supplemented with Complete EDTA-free Protease Inhibitor Cocktail (Roche) and 1 U/mL Benzonase (EMD Millipore). After clarification by high-speed centrifugation, the cell lysate was loaded onto a Strep-Tactin column (QIAGEN), and the bound protein was eluted using 10 mM desthiobiotin in Strep Buffer A. The eluted L-P complex was pooled and diluted with an equal volume of heparin buffer A (50 mM Tris-HCl, 10% glycerol, and 1 mM TCEP, pH 8.0), and was further purified using a heparin column (GE Healthcare). The L-P complex was eluted from the heparin column using a NaCl gradient to a final concentration of 500 mM. The protein was concentrated with an Amicon Ultra centrifugal filter, then loaded on to a size-exclusion column (Superpose 6 Increase 10/300, GE healthcare) equilibrated in 50 mM Tris-HCl, 500 mM NaCl, 10% glycerol, and 1 mM TCEP, pH 8.0. The fractions near the maximum height of the peak were combined, flash-frozen, and stored at À80 C.

Primer-extension reactions contained 10 nM RSV L-P complex, incubated with 200 nM RNA (5 0 -UUUGUUCGCGU) and 4 mM 5 0 biotin-labeled RNA primer (5 0 -biotin-pACGC) mixed in a buffer containing 20 mM Tris-HCl pH 7.5, 10 mM KCl, 6 mM MgCl 2 , 10% dimethyl sulfoxide (DMSO), 2 mM dithiothreitol (DTT), 10 mM GTP, 10 mM ATP, and 25 nM a 33 P-CTP. Each 10 mL reaction was incubated at 30 C for 2 hours and quenched with the addition of 0.1 M EDTA. Quenched reactions were transferred to a Flashplate (Perkin Elmer SMP103A001PK) and incubated for 1 hour at room temperature. The quenched reactions were then washed two times with 0.1% Tween-20, aspirated and read on a Microbeta Trilux (Perkin Elmer).

Each reaction contained 0.2 mM recombinant RSV L-P, 0.2 mM of an oligonucleotide template sequence derived from the RSV leader promoter (5 0 UUUGUUCGCGU 3 0 ) and 400 mM 5 0 -pACGC primer, mixed in a buffer containing 20 mM Tris-HCl pH 7.5, 10 mM KCl, 2 mM DTT, 0.01% Triton X-100, 10% DMSO, 0.2 U/mL RNasin (Ambion), and 6 mM MgCl 2 . Reactions were started by adding 100 nM a 33 P-GTP tracer with specific NTPs as described in the figure legend to a final volume of 10 mL, and incubated for 30 minutes at 30 C. Reactions were quenched by adding an equal volume of gel-loading buffer (Ambion). Samples were denatured at 95 C for 5 minutes and run on a 22.5% polyacrylamide urea sequencing gel for 1.5 hours at 80 W. The gel was dried, exposed to a phosphor screen, scanned on a Typhoon phosphorimager (GE Healthcare), and quantified using ImageQuant (GE Healthcare).

The methyltransferase activity was measured using a filter-binding assay, performed according to the method described previously . A 50 nM solution of RSV L-P complex was incubated with 1.8 mM purified synthetic RNA, 0.17 mM S-adenosyl methionine (SAM) and 0.8 mM 3 H-SAM (Perkin Elmer) in 50 mM Tris-HCl pH 8.0. After 3 h incubation at 30 C, reactions were quenched by a 20-fold dilution in cold water. Samples were transferred to DEAE filtermats (Perkin Elmer) using a Filtermat Harvester (Packard Instruments). The RNA-retaining mats were washed twice with 10 mM ammonium formate pH 8.0, twice with water and 

PHENIX: building new software for automated crystallographic structure determination

Real-space refinement in PHENIX for cryo-EM and crystallography

Taxonomy of the order Mononegavirales: update 2016

Intramolecular Fuzzy Interactions Involving Intrinsically Disordered Domains

The respiratory syncytial virus nucleoprotein-RNA complex forms a left-handed helical nucleocapsid

Structure and functional analysis of the RNA-and viral phosphoprotein-binding domain of respiratory syncytial virus M2-1 protein

Structural rearrangement of ebola virus VP40 begets multiple functions in the virus life cycle

RNA elongation by respiratory syncytial virus polymerase is calibrated by conserved region V

Generation of bovine respiratory syncytial virus (BRSV) from cDNA: BRSV NS2 is not essential for virus replication in tissue culture, and the human RSV leader region acts as a functional BRSV genome promoter

Biochemical characterization of the respiratory syncytial virus P-P and P-N protein complexes and localization of the P protein oligomerization domain

The structure of the RNA-dependent RNA polymerase from bovine viral diarrhea virus establishes the role of GTP in de novo initiation

Discovery of b-D-2 0 -deoxy-2 0 -a-fluoro-4 0 -a-cyano-5-aza-7,9-dideaza adenosine as a potent nucleoside inhibitor of respiratory syncytial virus with excellent selectivity over mitochondrial RNA and DNA polymerases

State of the Art in Respiratory Syncytial Virus Drug Discovery and Development

Transcription elongation factor of respiratory syncytial virus, a nonsegmented negative-strand RNA virus

The paramyxovirus polymerase complex as a target for next-generation anti-paramyxovirus therapeutics

ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps

Molecular Basis for the Selective Inhibition of Respiratory Syncytial Virus RNA Polymerase by 2 0 -Fluoro-4 0 -Chloromethyl-Cytidine Triphosphate

Biochemical Effect of Resistance Mutations against Synergistic Inhibitors of RSV RNA Polymerase

Activity of Oral ALS-008176 in a Respiratory Syncytial Virus Challenge Study

Coot: model-building tools for molecular graphics

Biochemical characterization of the respiratory syncytial virus N 0 -P complex in solution

Respiratory syncytial virus infection in elderly and high-risk adults

New antiviral approaches for respiratory syncytial virus and other mononegaviruses: Inhibiting the RNA polymerase. Antiviral Res

enzymes but with reduced activities

Identification and characterization of the binding site of the respiratory syncytial virus phosphoprotein to RNA-free nucleoprotein

Cytoplasmic inclusions of respiratory syncytial virus-infected cells: formation of inclusion bodies in transfected cells that coexpress the nucleoprotein, the phosphoprotein, and the 22K protein

Role of motif B loop in allosteric regulation of RNA-dependent RNA polymerization activity

Risk of primary infection and reinfection with respiratory syncytial virus

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

Structural basis for active site closure by the poliovirus RNA-dependent RNA polymerase

RNA replication by respiratory syncytial virus (RSV) is directed by the N, P, and L proteins

Structural basis for RNA-genome recognition during bacteriophage Qb replication

The product of the respiratory syncytial virus M2 gene ORF1 enhances readthrough of intergenic junctions during viral transcription

Effectiveness of Palivizumab in Preventing RSV Hospitalization in High Risk Children: A Real-World Perspective

Structure of the SARS-CoV nsp12 polymerase bound to nsp7 and nsp8 co-factors

Transformer proteins

Solution and crystallographic structures of the central region of the phosphoprotein from human metapneumovirus

A conserved motif in region v of the large polymerase proteins of nonsegmented negative-sense RNA viruses that is essential for mRNA capping

Structure of the L Protein of Vesicular Stomatitis Virus from Electron Cryomicroscopy

Inhibitors of respiratory syncytial virus replication target cotranscriptional mRNA guanylylation by viral RNA-dependent RNA polymerase

Structural analysis of the human respiratory syncytial virus phosphoprotein: characterization of an a-helical domain involved in oligomerization

Structural properties of the human respiratory syncytial virus P protein: evidence for an elongated homotetrameric molecule that is the smallest orthologue within the family of paramyxovirus polymerase cofactors

Biochemical properties of hepatitis C virus NS5B RNA-dependent RNA polymerase and identification of amino acid sequence motifs essential for enzymatic activity

How order and disorder within paramyxoviral nucleoproteins and phosphoproteins orchestrate the molecular interplay of transcription and replication

Identification of temperature-sensitive mutations in the phosphoprotein of respiratory syncytial virus that are likely involved in its interaction with the nucleoprotein

Mechanism for coordinated RNA packaging and genome replication by rotavirus polymerase VP1

Increased genetic and phenotypic stability of a promising live-attenuated respiratory syncytial virus vaccine candidate by reverse genetics

Interaction between human respiratory syncytial virus (RSV) M2-1 and P proteins is required for reconstitution of M2-1-dependent RSV minigenome activity

The Interferon Type I/III Response to Respiratory Syncytial Virus Infection in Airway Epithelial Cells Can Be Attenuated or Amplified by Antiviral Treatment

Signature motifs of GDP polyribonucleotidyltransferase, a non-segmented negative strand RNA viral mRNA capping enzyme, domain in the L protein are required for covalent enzyme-pRNA intermediate formation

Factors affecting de novo RNA synthesis and back-priming by the respiratory syncytial virus polymerase

Intrinsic Disorder to Order Transitions in the Scaffold Phosphoprotein P from the Respiratory Syncytial Virus RNA Polymerase Complex

X-ray structure and activities of an essential Mononegavirales L-protein domain

New Insights into Structural Disorder in Human Respiratory Syncytial Virus Phosphoprotein and Implications for Binding of Protein Partners

UCSF Chimera-a visualization system for exploratory research and analysis

Identification of four conserved motifs among the RNA-dependent polymerase encoding elements

cryo-SPARC: algorithms for rapid unsupervised cryo-EM structure determination

Molecular architecture of the vesicular stomatitis virus RNA polymerase

Critical phosphoprotein elements that regulate polymerase architecture and function in vesicular stomatitis virus

Deciphering key features in protein structures with the new ENDscript server

CTFFIND4: Fast and accurate defocus estimation from electron micrographs

Alignment of cryo-EM movies of individual particles by optimization of image translations

Electron microscopy: Ultrastable gold substrates for electron cryomicroscopy

Point mutations which drastically affect the polymerization activity of encephalomyocarditis virus RNA-dependent RNA polymerase correspond to the active site of Escherichia coli DNA polymerase I

The Structure of the Human Respiratory Syncytial Virus M2-1 Protein Bound to the Interaction Domain of the Phosphoprotein P Defines the Orientation of the Complex

Assessing Variation in the Cost of Palivizumab for Respiratory Syncytial Virus Prevention in Preterm Infants

Global, regional, and national disease burden estimates of acute lower respiratory infections due to respiratory syncytial virus in young children in 2015: a systematic review and modelling study

Structural analysis of human respiratory syncytial virus p protein: identification of intrinsically disordered domains

Fine mapping and characterization of the L-polymerase-binding domain of the respiratory syncytial virus phosphoprotein

Molecular insights into replication initiation by Qb replicase using ribosomal protein S1

RNA synthesis in a cage-structural studies of reovirus polymerase lambda3

Crystal structure of a nucleocapsid-like nucleoprotein-RNA complex of respiratory syncytial virus

Palivizumab, a humanized respiratory syncytial virus monoclonal antibody, reduces hospitalization from respiratory syncytial virus infection in high-risk infants

The nine C-terminal amino acids of the respiratory syncytial virus protein P are necessary and sufficient for binding to ribonucleoprotein complexes in which six ribonucleotides are contacted per N protein protomer

The respiratory syncytial virus M2-1 protein forms tetramers and interacts with RNA and P in a competitive manner

MonoRes: Automatic and Accurate Estimation of Local Resolution for Electron Microscopy Maps

Subunit I of G beta replicase and 30 S ribosomal protein S1 of Escherichia coli. Evidence for the identity of the two proteins

Discovery of 4 0 -chloromethyl-2 0 -deoxy-3 0 ,5 0 -di-O-isobutyryl-2 0 -fluorocytidine (ALS-8176), a first-in-class RSV polymerase inhibitor for treatment of human respiratory syncytial virus infection

Addition of a missense mutation present in the L gene of respiratory syncytial virus (RSV) cpts530/1030 to RSV vaccine candidate cpts248/404 increases its attenuation and temperature sensitivity

Functional cDNA clones of the human respiratory syncytial (RS) virus N, P, and L proteins support replication of RS virus genomic RNA analogs and define minimal trans-acting requirements for RNA replication

They were then soaked with scintillation fluid (Perkin Elmer), and 3 H-methyl transfer to the RNA substrates was determined using a Wallac MicroBeta TriLux Liquid Scintillation Counter

After RNA elongation with 2 0 O-pivaloyloxymethyl phosphoramidite monomers (Chemgenes, USA), the 5 0 -hydroxyl group was phosphorylated and the resulting H-phosphonate derivative was oxidized and activated into a phosphoroimidazolidate derivative to react with pyrophosphate (pppRNA) or guanosine diphosphate (GpppRNA)

Cryo-EM data collection A cryo-EM dataset of the RSV L-P complex was collected on a Titan Krios operating at 300kV and equipped with a K2 Summit detector. An initial dataset was collected on samples frozen in 1.2/1.3 holey carbon grids

Electron Microscopy Sciences) were plasma cleaned for 30 s using a Gatan Solarus 950 with a 4:1 O 2 :H 2 ratio. A 0.57 mg/mL solution of L-P complex in 50 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM TCEP and 10% glycerol was diluted with an equal volume of 20 mM Tris-HCl pH 8.0, 200 mM NaCl immediately before 3 mL was deposited onto grids and plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Scientific) set to 100% humidity and 4 C, with a wait time of 0.5 s, a blot time of 4 s and a blot force of 1. Data were collected at 22,500x magnification, corresponding to a calibrated pixel size of 1.075 Å . A total of 30 frames were collected for each micrograph

Template-based picking identified 1,455,759 particles, which was reduced to 622,521 particles after 2D classification and 241,669 particles after two iterative rounds of heterogeneous 3D refinement. Local motion correction was performed on the final particle stack (Rubinstein and Brubaker, 2015), followed by homogeneous 3D refinement, which resulted in a 3.4 Å map. Upon initial building of the L protein model and docking of the closely related hMPV P oligomerization domain structure in our map, it was possible to identify the portions of the map that corresponded to the P protein monomers

Sequence analysis was carried out to check the integrity of all the constructs. Cells at 90% confluence in 48-well dishes were transfected with a plasmid mixture containing 125 ng of pM/Luc, 125 ng of pN, 125 ng of pP, 62.5 ng of pL, and 31 ng of pM2-1, as well as 31 ng of pRSV-b-Gal (Promega) to normalize transfection efficiencies