key: cord-0267050-8s9i71lq
authors: Xie, Jin; Wang, Li; Zhai, Guanglei; Wu, Daitze; Lin, Zhaohu; Wang, Manfu; Yan, Xiaodong; Gao, Lu; Huang, Xinyi; Fearns, Rachel; Chen, Shuai
title: Structural architecture of a dimeric paramyxovirus polymerase complex
date: 2021-09-13
journal: bioRxiv
DOI: 10.1101/2021.09.13.460081
sha: 1ea87752fbfb84df219631f6af4e5313ea76a3f7
doc_id: 267050
cord_uid: 8s9i71lq

Human parainfluenza virus type 3 (hPIV3), a member of non-segmented, negative-strand RNA viruses (nsNSVs), is the second most common cause of severe respiratory disease in pediatrics. The transcription and replication processes of nsNSVs are catalyzed by a multi-functional RNA-dependent RNA polymerase (RdRp) complex composed of the large protein (L) and the phosphoprotein (P). Previous studies have shown that the polymerase can adopt a dimeric form, however, the structure of the dimer and how it functions are not understood. Here we determined the cryo-EM structure of hPIV3 L-P complex at 2.7 Å with substantial structural details. A putative catalytic magnesium ion could be built in our structure, and structural comparison revealed atomic features conserved with other RNA viruses. Interactions identified between the two priming and intrusion loops and the connector domain potentially trigger the spatial movement of three C-terminal L domains for different steps of transcription and replication. Structural comparison with other nsNSV RdRps suggests common features of L-P binding. Furthermore, we report for the first time the structural basis of the L-L interaction in the partially modelled dimeric L-P structure, in which the connector domain of one L is positioned at the putative RNA template entry of the other L. Based on these findings, we propose a model by which L dimerization promotes efficient conversion of nascent RNA into a template.

The non-segmented negative-strand RNA viruses (nsNSVs) include numerous human pathogens, such as 29 respiratory syncytial virus (RSV), human parainfluenza viruses (hPIVs) and rabies virus (RABV) 1 . HPIV3, 30 one of four hPIV subtypes, is the second most common cause of severe viral respiratory diseases in infants 31 and children 2,3 . The RNA genome of nsNSVs is packaged by the nucleoprotein (N) into a helical 32 ribonucleoprotein (RNP) complex that acts as the template for genome transcription and replication by the 33 initiation and elongation mechanism of nsNSV polymerase, and reveals the structural conservation of L-P 48

binding. More interestingly, the L-L interaction was directly observed and suggests a novel model of nsNSV 49 genome replication. 50 51 2.7 Å cryo-EM structure of hPIV3 L-P polymerase complex 52 Full-length human PIV3 L and P proteins (Fig. 1a) were co-expressed in Sf21 cells. The purified L-P 53 complex ( Fig. 1b) showed strong binding to an RNA duplex (Extended Data Fig. 3a ) and its RdRp activity 54 was verified by a fluorescence-based primer extension assay 25 (Fig. 1c) . 55

We employed single-particle cryo-EM to solve the hPIV3 L-P complex structure at a resolution up to 2.7 56 Å (Fig. 1d , e, Extended Data Fig. 1 , Extended Data Table 1 ). This is the first nsNSV RdRp structure with 57 a resolution higher than 3.0 Å, allowing us to build a more detailed atomic model. All five domains of L 58 protein and four copies of OD domain and single XD domain of P protein were built into the density map. 59

The overall structural architecture is similar to that of canine PIV5 L-P 11 . Interestingly, a large extra blob 60 of electron density, not reported in previous L-P structures, was observed near the intact L protein and was 61 successfully assigned as the CD domain of a second L protein (Fig. 1d , e, Extended Data Fig. 1 ). In addition 62 to the partially modelled dimeric L-P structure, another main class of particles lacked density for the second 63 CD domain and was reconstructed to a 3.3 Å resolution structure representing the monomeric L-P complex 64 ( Fig. 1f , g, Extended Data Fig. 1 ). Since these two structures show nearly the same arrangement except for 65 the second CD domain, we used the 2.7 Å structure for further analysis. 66

Consistent with previously determined RNA polymerase structures 5,10,11,17,26-28 (Extended Data Fig. 2) , the 69

RdRp domain of hPIV3 L-P complex folds into the canonical "right hand" fingers-palm-thumb subdomains, 70 while the catalytic active site is composed of seven conserved motifs (A-G) (Fig. 2a, b) . Motifs A-E are 71 located in the palm subdomain, while motifs F and G are located in the finger subdomain. In order to further 72 understand the RdRp active site of nsNSVs, we compared our structure to the RNA/Mg 2+ -bound RdRp 73 structures of influenza B virus (FluB) 28 and SARS-CoV-2 27 , as the representatives of segmented negative-74 strand RNA viruses (sNSVs) and positive-strand RNA viruses, respectively. The hPIV3 L RdRp domain 75

shows similar structural architecture to that of FluB and SARS-CoV-2 (r.m.s.d of 2.93 Å over 349 Cα atoms 76 and 3.43 Å over 319 Cα atoms, respectively) (Extended Data Fig. 2) . Furthermore, motifs A-G could be 77 well overlaid, and the proposed catalytic residues 772-GDN-774 of hPIV3 RdRp could be also 78 superimposed with FluB (443-SDD-445) and SARS-CoV-2 (759-SDD-761) (Fig. 2c, d) . One magnesium 79 ion at the catalytic center could be built in our hPIV3 structure due to the well-resolved electron density 80 Fig. 1f ), which has not been reported in previous nsNSV L structures. The 81 presumed catalytic Mg 2+ is located at the similar position as one of the two Mg 2+ present in FluB and SARS-82

CoV-2 structures, and coordinated by the side chain oxygen of the catalytic residue Asp773 and the main 83 chain oxygen of motif A residue Leu664 (Fig. 2c, d) . Similar to FluB 28 and SARS-CoV-2 27 , hPIV3 L 84 appears to possess the conserved motif F residues Arg552 and Lys543/Phe554 to stabilize the incoming 85 nucleotide and the template strand RNA at the +1 site, respectively (Fig. 2c, d) . In addition, motif G directs 86 the template strand RNA into the active site with a conserved positively charged residue 27, 28 , Lys475 for 87 hPIV3 (Fig. 2c, 

The PRNTase domain of L is multifunctional, facilitating de novo initiation via a priming loop, capping 94 the nascent viral mRNAs, and regulating RNA elongation 33-38 (Fig. 2e, Extended Data Fig. 4) . Compared 95 to the relatively rigid structure of the core RdRp-PRNTase domains, the three C-terminal CD-MTase-CTD 96 domains of L behave more dynamically 11,24 (Extended Data Fig. 1c-e) . The spatial position of the hPIV3 97 CD-MTase-CTD relative to RdRp-PRNTase is different from the reported nsNSV structures 10,11,15-19 (Fig.  98 2f, Extended Data Fig. 5) . A short β-strand extended from the hPIV3 RdRp domain forms a β-sheet, aiding 99 the positioning of CTD. Comparison of the different structures suggests that they represent the RdRp at 100 different stages of RNA synthesis. 101

In the reported VSV and RABV structures, the priming loop reaches towards the central cavity, 102

representing a pre-initiation state 10,15,16 . In contrast, the hPIV3 priming loop (Val1243-Ser1259) retracts 103 considerably from the central tunnel and props against CD, leading to the slight shift of CD and subsequent 104 movement of MTase-CTD. Instead of the priming loop, an intrusion loop (Thr1280-Ser1305), containing 105 the catalytic HR motif for capping, partially projects out into RdRp cavity, a feature also observed in PIV5 106 L 11 (Fig. 2e , Extended Data Fig. 4a -d, 5a-d). Thus, both hPIV3 and PIV5 structures appear to be in a post-107 initiation state, but have several differences. In the hPIV3 structure, the distinct positioning of the HR motif 108 towards the RNA cavity and the fact that the intrusion loop could accommodate extension of several base- More detailed L-P interactions are observed in our structure than in PIV5 11 (Fig. 3a-h) . Tetrameric OD 124 domains of hPIV3 P constitute a long helical bundle bound to the RdRp domain of L with longer visible 125 loops than PIV5 (Fig. 3i) . Each of the four P monomers (P1-P4) adopts asymmetric conformations. Two 126 proximal subunits P1 (Arg451-Arg470) and P4 (Leu464-Lys473) make extensive contacts with L ( Fig.  127 3a-f). Residues Asn463-Glu469 of P4 are liberated from the long helix, and residues Lys465-Met467 form 128 a typical antiparallel β-sheet with Gln387-Lys389 of L (Fig. 3e ). In addition, the neighboring residue 129

Phe390 along with Ile452 and Leu678 of L inserts into the exposed hydrophobic core composed of P1 and 130 P4 (Fig. 3d) . Similar β-sheet and hydrophobic interactions are also present in RSV and hMPV structures 17-131

. Among the three C-terminal α helices (named α1-α3 here) of hPIV3 P-XD, the α1 and its upstream loop, 132 corresponding to the single α-helix at the C-terminus and the neighboring linker observed in RSV and 133 hMPV L-P structures 17-19 , contribute to the majority of interactions between P-XD and L (Fig. 3g, h) . 134

Most of the residues involved in the L-P interactions are highly conserved in the closely related 135 paramyxoviruses such as caprine PIV3 (cPIV3), human PIV1 (hPIV1) and Sendai virus (SeV), while less 136 conserved in the distantly related paramyxovirus PIV5 and other nsNSVs (Extended Data Figs. 8, 9) . 137

However, both OD and XD domains of hPIV3 and PIV5 P proteins adopt similar conformations and bind 138

to the same approximate positions on L (Fig. 3i) . Interestingly, RSV 17,18 and hMPV 19 also dock the 139 tetrameric helical OD domains of P on the L surface like hPIV3 with some similar interaction features 140 described as above, and extend their C-terminal helices of one monomer to roughly the same position as 141 hPIV3 (Fig. 3j, k) . In addition, linker density was observed between P4-OD and the C-terminal helices of 142 P-XD in PIV5 11 , comparable to residues 163-210 of RSV P4. In summary, elements of L-P binding seem 143 to be conserved between the Paramyxoviridae and Pneumoviridae families despite diversity in P sequence 144 and structure. forms cation-π interaction with Arg1101 of PRNTase (Fig. 4d ). CD-2 has a similar overall structure as the 160 CD domain of the intact L, except for the two loops that contribute to the L-L interface and the bent helix 161 α2 of CD-2 (Extended Data Fig. 7a ). It reveals that the two CD domains can bind to distinct interfaces on 162 Previous studies have shown that inactive SeV L variants with mutations in different domains could 183 complement each other to restore mRNA transcription and genome replication albeit at a low level, and 184 complementation for leader RNA synthesis is efficient, indicating that the dimerization or oligomerization 185 of L-P polymerase might be required for transcription and replication 21 . Based on these findings and the 186 captured asymmetric dimer of hPIV3 L-P, we proposed a dimeric L-P polymerase model for nsNSV RNA 187 replication (Fig. 4g) , although we cannot exclude the possibility that nsNSV transcription also needs 188 dimeric polymerase. P-XD first binds to N protein in the RNP complex and scans along the genome 6 . The 189 3′ terminus of the template RNA is loaded onto the central RNA tunnel to initiate RNA synthesis. The (PDB 6PZK) and hMPV (PDB 6U5O), P4 that extends its C-terminal to the place roughly the same to P-348

XD of hPIV3 is colored in black, while another three P copies are colored in grey. 349 Protein purification of hPIV3 L-P complex 376 The pellet of Sf21 cells expressing the hPIV3 L-P complex was resuspended through high-pressure 377 homogenizer in the lysis buffer (50 mM Tris, pH 7.5, 500 mM NaCl, 10% glycerol, 6 mM MgSO 4 , 1 mM 378 dithiothreitol (DTT), 1% Triton X-100) supplemented with cOmplete™, EDTA-free protease inhibitor 379 cocktail (Roche). After high-speed centrifugation at 5,0000× g for 60 min at 4°C, the supernatant containing 380 the target proteins was loaded onto anti-FLAG G1 affinity resin (Genscript). The resin was washed using 381 buffer A (20 mM Tris, pH 7.5, 300 mM NaCl, 5% glycerol, 6 mM MgSO 4 , 1 mM DTT), and the bound 382 proteins were eluted using 0.2 mg/ml FLAG peptide in buffer A. The eluted proteins were further purified 383 using a size-exclusion column (Superose 6 Increase 10/300 GL, GE healthcare) equilibrated with buffer B 384 (20 mM Tris, pH 7.5, 300 mM NaCl, 1% glycerol, 6 mM MgSO 4 , 1 mM DTT). The peak fractions were 385 collected and analyzed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE). 386

Fractions that contained the L-P complex were combined and concentrated to 1.2 mg/ml. The final sample 387 was flash-frozen, and stored at −80 °C. The protein homogeneity was characterized by negative-stain EM. against hPIV3. a, The binding affinity of hPIV3 L-P complex to the immobilized RNA duplex was 570 determined by SPR assay. b, Enzymatic assay showed that the ATP analog RDV-TP has a competition 571 effect on ATP. c, IC 50 determination of RDV-TP against hPIV3 L-P complex. d, e, Fluorescence-based 572 plaque reduction assay was used to determine the EC 50 of remdesivir (RDV) against hPIV3 in HEp2 cells.

All experiments were tested in triplicate. 

Taxonomic update of phylum Negarnaviricota (Riboviria: Orthornavirae), 200 including the large orders Bunyavirales and Mononegavirales. Archives of Virology

Evaluation of combined live, attenuated respiratory syncytial virus and 203 parainfluenza 3 virus vaccines in infants and young children

Epidemiology and clinical presentation of the four human parainfluenza virus 206 types

Initiation and regulation of paramyxovirus transcription and replication. 208 Virology 479-480

Structural insights into RNA polymerases of 210 negative-sense RNA viruses

Paramyxovirus RNA synthesis, mRNA editing, and genome hexamer phase: A 212 review

Near-atomic cryo-EM structure of the helical measles virus nucleocapsid

Synthesis of message and genome RNAs in vitro by Sendai virus-infected cell 216 nucleocapsids

Structures of the Mononegavirales polymerases

Structure of the L protein of vesicular stomatitis virus from electron 219 cryomicroscopy

Structure of a paramyxovirus 221 polymerase complex reveals a unique methyltransferase-CTD conformation

Polymerases of paramyxoviruses and pneumoviruses

Rescue of Sendai virus from viral 226 ribonucleoprotein-transfected cells by infection with recombinant vaccinia viruses carrying Sendai 227 virus L and P/C genes

Complexes of Sendai virus NP-P and 229 P-L proteins are required for defective interfering particle genome replication in vitro

Structure of the vesicular stomatitis virus L protein in complex with its 232 phosphoprotein cofactor

Structure of a rabies virus polymerase 234 complex from electron cryo-microscopy

Structure of the respiratory syncytial virus polymerase complex

Automated acquisition of cryo-electron micrographs for single particle 510 reconstruction on an FEI Tecnai electron microscope

RELION: Implementation of a Bayesian approach to cryo-EM structure 513 determination

Algorithms for rapid 515 unsupervised cryo-EM structure determination

UCSF Chimera -A visualization system for exploratory research and 517 analysis

Coot: Model-building tools for molecular graphics

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

MolProbity: All-atom structure validation for macromolecular crystallography

Deciphering key features in protein structures with the new ENDscript 525 server

Inference of macromolecular assemblies from crystalline state

reduction titers were calculated by regression analysis of the dilution of remdesivir compared to control 428 wells incubated with 1% DMSO. 429 Surface plasmon resonance (SPR) assay 430 The binding affinity between the purified hPIV3 L-P complex protein and a RNA duplex were measured at 431 room temperature using a Biacore S200 system (GE Healthcare) with a SA chip (Cytiva, BR100531). The 432RNA duplex with the same sequence as that used in the primer extension assay except for 5′-biotin at the 433 primer strand was immobilized on the chip. A blank channel of the chip was used as the negative control. 434The running buffer contained 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.05% surfactant P20 and 6 mM 435MgCl 2 . Proteins were serially diluted to 1.95~250 nM using the running buffer and then loaded to flow 436 through the chip surface. After binding and dissociation of each sample, the chip was regenerated using 50 437 mM NaOH with a short contact time of 5s. The sensorgrams were analyzed using the Biacore S200 438Evaluation Software (version1.1) with 1:1 kinetics binding model. 439

An aliquot of 3.5 μL of hPIV3 L-P complex at 0.8 mg/ml was applied to a freshly glow-discharged 441 Quantifoil R1.2/1.3 300-mesh grid. The sample was immediately blotted at 4°C and 100% relative humidity, Patch-CTF, a total of 5693 micrographs were selected for subsequent processing. To generate templates for 454 automatic particle picking, 642 micrographs were selected, and 589,198 particles were auto-picked using 455 cryoSPARC's blob picker and extracted with a box size of 128 pixels after binning. After 2D classification, 456 100,387 particles were selected for 3D classification using Ab-Initio reconstruction in cryoSPARC, and 457 three classes were generated as the initial reference models. Then 50 2D-templates were projected from the 458 model with clear structural features. For the dataset of hPIV3 L-P complex, 4,511 micrographs were 459 selected based on fitted resolution better than 4 Å, and a total of 3,288,374 particles were picked using 460 templates generated previously and extracted with a box size of 150 pixels after binning. 636,487 particles 461 were selected after two rounds of 2D classification based on the complex integrity. Then heterogeneous 462 refinement was performed using previously generated Ab-Initio models. A subset of 373,483 particles from 463 the class showing clear structural features was selected and re-extracted with a box size of 360 pixels 464 without binning, and the resolution reached 3.3 Å after homogeneous refinement. For further classification, 465 the full complex model and two erased models were used as 3D volume templates for heterogeneous 466 refinement. The two classes of high quality (class 1 and 2, with and without a large extra blob of electron 467 density, respectively) were subjected to homogeneous refinement, local refinement and non-uniform 468 refinement using cryoSPARC. Then, CTF refinement and Bayesian polishing refinement were performed 469 in RELION 47 . Finally, we obtained the 2.7 and 3.3 Å cryo-EM density maps for class 1 and class 2, 470respectively. In addition, several 2D classes generated from 46, 722 particles had a bigger size and appeared 471 to be dimeric complex, but we failed to reconstruct further for these classes. 472

The canine PIV5 L-P structure (PDB entry: 6V85) 11 was used to guide the building of the atomic model of 474 the hPIV3 L-P class 1 and 2. The complex of RdRp-PRNTase domains of L and four copies of P-OD and 475 single P-XD of PIV5 was placed and rigid-body fitted well into the class 1 cryo-EM map using UCSF 476 Chimera 49 . The CD and MTase-CTD domains of L were rigid-body docked separately with some rotation. 477Manual model building was carried out using Coot 50 and refinement of the coordinates was performed using 478 phenix.real_space_refine 51 . One magnesium ion at the catalytic center could be built due to its well-resolved the full L protein residues except for the N-terminal 7 residues, Tyr611-Lys637, Leu1292-Met1299, 488Ile1693-Asp1706, Thr1745-Thr1762 and Thr2095-Lys2113; the P protein residues with four copies of OD 489 domains (Asp435-Gly471, Ala434-Met467, Asp435-Gly472 and Asp435-Asp475 of subunits P1, P2, P3 490 and P4, respectively) and single XD domain (Asn539-Gln603); and the CD domain of second L protein 491 residues Asp1450-Leu1483 and Ile1495-Ile1712. The final hPIV3 L-P model of class 2 comprises nearly 492