A Fijivirus Major Viroplasm Protein Shows RNA-Stimulated ATPase Activity by Adopting Pentameric and Hexameric Assemblies of Dimers

ABSTRACT Fijiviruses replicate and package their genomes within viroplasms in a process involving RNA-RNA and RNA-protein interactions. Here, we demonstrate that the 24 C-terminal residues (C-arm) of the P9-1 major viroplasm protein of the mal de Río Cuarto virus (MRCV) are required for its multimerization and the formation of viroplasm-like structures. Using an integrative structural approach, the C-arm was found to be dispensable for P9-1 dimer assembly but essential for the formation of pentamers and hexamers of dimers (decamers and dodecamers), which favored RNA binding. Although both P9-1 and P9-1ΔC-arm catalyzed ATP with similar activities, an RNA-stimulated ATPase activity was only detected in the full-length protein, indicating a C-arm-mediated interaction between the ATP catalytic site and the allosteric RNA binding sites in the (do)decameric assemblies. A stronger preference to bind phosphate moieties in the decamer was predicted, suggesting that the allosteric modulation of ATPase activity by RNA is favored in this structural conformation. Our work reveals the structural versatility of a fijivirus major viroplasm protein and provides clues to its mechanism of action.

P lant diseases caused by fijiviruses (family Spinareoviridae and order Reovirales) severely threaten crop production. The Mal de Río Cuarto virus is a member of the genus Fijivirus (1), which causes the most severe and economically important maize viral disease in Argentina (2), one of the largest producing and exporting nations worldwide (3). Delphacid planthopper insects transmit the virus in a persistent propagative manner (4). Other fijiviruses severely affect rice and maize production in Asia and Europe (5).
Reovirids replicate and assemble within membraneless cytoplasmic inclusion bodies called viroplasms or viral factories. These structures are formed early during infection and are composed of viral proteins and RNA as well as several host factors (6). Fijivirus particles contain 10 double-stranded RNA (dsRNA) genomic segments that encode at least 12 proteins (1). Viroplasms produced by the fijiviruses rice black-streaked dwarf virus (RBSDV) and southern rice black-streaked dwarf virus (SRBSDV) present two distinct morphologies, one granular (predominantly composed by the nonstructural viral protein P9-1) and another filamentous (predominantly composed by the nonstructural protein P5) (7,8). The nonstructural protein P6 is driven to both types of viroplasms by direct protein-protein interactions with P9-1 and P5 (7,9). Viroplasms were shown to be highly dynamic. Viral RNA has been proposed to accumulate in the granular viroplasm, whereas viral progeny core and complete virus particles tend to accumulate in the more electrodense filamentous viroplasm (7).
A few viroplasm proteins from animal and plant reoviruses have been structurally characterized. Studies on rotavirus nonstructural protein NSP2 have shown that it works as a doughnut-shaped octamer with a central pore and prominent diagonal grooves where NSP5 and single-stranded RNA (ssRNA) bind (21,22). In structural proximity to the RNA-binding grooves, each NSP2 monomer has clefts containing an NTPase active site (23). In turn, the crystallographic structure of the N-terminal domain of bluetongue virus (BTV; genus Orbivirus) NS2 showed that this protein homomultimerizes through extensive monomer-monomer interactions (24), and electron microscopy studies of the fulllength version of the protein revealed that the oligomers have a ring-like shape (25).
Regarding plant reovirids, cryo-electron microscopy (cryo-EM) analysis of Pns9 from rice gall dwarf virus (RGDV; genus Phytoreovirus) revealed the formation of octamers with an internal pore (26). Similarly, the crystallographic structures of RBSDV and SRBSDV P9-1 showed that these proteins form dimers that interact with each other through C-terminal regions of 24 residues (C-arms), giving rise to cylindrical octamers (27,28). In both proteins, the deletion of the C-arm prevents multimerization (27,29,30), whereas in RBSDV P9-1, it hinders the formation of VLSs in insect cells (27). Consistent with these findings, we have previously shown that the MRCV P9-1 C-terminal half (residues 155 to 337) is required for VLS formation in insect cells (16) and that the deletion of the C-arm (residues 314 to 337) affects its self-interactions in yeast two-hybrid assays (18). The structure and function of viroplasm components underpin the precise coordination of virus replication and packaging. These steps are particularly complex in the case of viruses with segmented dsRNA genomes that require equimolar packaging of all segments. The mechanisms underlying this process are being increasingly understood in animal reoviruses where phosphorylation cascades on rotavirus NSP2 and NSP5 and the RNA chaperone function of NSP2 have crucial roles, as recently reviewed (6,31). However, much less is known about the structural and functional aspects of the viroplasms from plant-infecting reoviruses.
To shed light on MRCV viroplasm assembly and function, we performed an integrative structural characterization of P9-1 showing that C-arm-driven oligomerization leads to quaternary structural conformations with an RNA-boosted ATPase activity previously unidentified for a major viroplasm protein of the Reovirales order. These findings may have significant impacts in the design of antiviral strategies for plant disease control in important crops.

RESULTS
The P9-1 C-arm is required for the formation of VLSs in rice protoplasts and insect cells. We have previously shown that P9-1 forms VLSs in the cytoplasm of both plant and insect cells (16)(17)(18). To assess the contribution of the P9-1 C-arm in the formation of such structures, we transiently expressed P9-1 (337 residues) and P9-1DC-arm (lacking residues 314 to 337) fused to the green fluorescent protein (GFP) and analyzed their subcellular localization in rice protoplasts and insect Sf9 cells by confocal imaging. As expected, GFP:P9-1 fluorescence was located in punctate, distinct cytoplasmic inclusion bodies both in plant and insect cells (Fig. 1). Deletion of the C-arm resulted in a dispersed cytoplasmic GFP fluorescence in both systems ( Fig. 1), indicating that VLS formation was impaired. These results suggest that the C-arm plays a key role during P9-1 multimerization, which is required for VLS formation.
P9-1 multimerizes into high-molecular-mass complexes that rely on the presence of the C-arm. To evaluate whether VLS formation is a result of P9-1 self-interactions leading to multimerization, we subsequently analyzed P9-1 and P9-1DC-arm oligomeric states in solution by size-exclusion chromatography (SEC) coupled to static light scattering (SLS).
Both proteins were produced recombinantly in bacteria and purified by immobilized metal affinity chromatography (IMAC) followed by SEC, and their purity was assessed by SDS-PAGE ( Fig. 2A). Under reducing conditions, the proteins migrated according to the theoretical molecular mass (MM) of their monomeric species (full-length P9-1, 44.9 kDa; P9-1DC-arm, 37.2 kDa). The SEC-SLS analyses showed that P9-1 and P9-1DC-arm mostly eluted as oligomeric structures harboring ;10.4 protomers (experimental MM = 467.9 6 32.7 kDa) (Fig. 2B) and ;2.3 protomers (experimental MM = 85.3 6 5.8 kDa) (Fig. 2C), respectively. The P9-1DC-arm sample was characterized by a persistent high nonspecific SLS signal (with low refractive index), which suggests the formation of soluble aggregates possibly due to the instability of the protein construct. Overall, these results indicate that P9-1 behaves as a higher-order oligomer, while the P9-1DC-arm construct does not multimerize beyond a dimeric state. These findings are in agreement with those previously reported for RBSDV P9-1, where the C-arm is required for octamer formation but not for the dimer assembly (27).
To provide further evidence on the oligomerization states of the two proteins, dynamic light scattering (DLS) measurements were performed. P9-1 showed a major population obtained by number distribution corresponding to a hydrodynamic  S1A and B), which is in very good agreement with the estimations based on the SEC-SLS experiments. The D h obtained via analytical SEC for P9-1 (;14.6 nm) was also consistent with the value determined via DLS (Fig. S1C). Importantly, these results were also in agreement with the SEC profile and the D h parameter obtained for full-length P9-1 expressed in insect Sf9 cells ( Fig. S2; Table S1). Taken together, these results consistently indicated that P9-1 expressed in prokaryotic and eukaryotic systems forms higher-order oligomers with stoichiometries that likely exceed an octameric assembly (theoretical MM of 359.2 kDa), as reported for RBSDV P9-1 (27).
Considering that most of the structures described for P9-1 homologous proteins and our previous works describing P9-1 functional properties were obtained following expression in bacteria, we subsequently pursued the structural characterization of P9-1 using this protein source.
The crystal structure of P9-1DC-arm reveals a dimeric arrangement. Attempts to crystallize the full-length P9-1 protein were unsuccessful since they consistently rendered low-quality crystals. Instead, P9-1DC-arm crystallized in the tetragonal space group P4 3 2 1 2 with unit cell parameters of a = b = 86.56Å and c = 95.60Å, and the best diffraction data set was collected to a maximum resolution of 3.47 Å ( Table 1). The crystal structure was solved by the molecular replacement method using the atomic coordinates of RBSDV P9-1 as a search model (Protein Data Base [PDB] code: 3VJJ), where one independent molecule of P9-1DC-arm was found in the asymmetric unit.  The final 2mF o -DF c electron density map was consistent, with no chain breaks for most of the protein backbone, except for the initial four N-terminal residues and the regions comprising the residues 20 to 43, 71 to 72, 108 to 110, 131 to 154, 229 to 237, and 265 to 268, which correspond mainly to loops. Despite the moderate resolution reached, the final refined model showed good stereochemistry parameters (98th percentile according to MolProbity score [32] on structures of comparable resolution) and acceptable refinement statistics (R work = 0.22 and R free = 0.28) ( Table 1).
The structure of P9-1DC-arm bears nine a-helices (aI to aIX) and nine b-strands (bA to bI) (Fig. 3A). The longest helix aIV crosses the entire protein fold enclosed by  the other a-helices forming a compact helix bundle. Three antiparallel stranded b-sheets constituted by the strands bA(;),bB(:) (b-sheet 1), bC(:),bD(;),bE(:),bI(;) (b-sheet 2), and bF(;),bG(:),bH(;) (b-sheet 3) are exposed to the solvent flanking a side of the helix bundle almost perpendicular with respect to the helix aIV (Fig. 3A). The strands bF and bG from b-sheet 3 protrude out from the global protein scaffold as a b-hairpin, while the loops bA-bB  and bD-bE (126 to 155) are partially defined by the electron density map, revealing local flexibility (Fig. 3A). These observations are consistent with the prediction of intrinsically disordered regions (IDRs) based on the P9-1 amino acid sequence (Fig. S3). Remarkably, the loop bA-bB comprises the RNA binding site previously described for RBSDV P9-1 (residues 25 to 44), while the loop bD-bE contains the PEST motif (KTESTSSELPAK, residues 142 to 153) for putative proteasome-mediated degradation (Fig. S4). The P9-1DC-arm folds similar to RBSDV P9-1 (27), as revealed by a root mean square deviation (RMSD) of 1.21 Å for 221 aligned C a atoms (Fig. 3B). However, some appreciable differences are noted. The protruding hairpin bF-bG is not defined in RBSDV P9-1, while the loop bE-aIV (160 to 172) in the P9-1DC-arm structure shows an altered scaffold mainly due to the absence of an a-turn (Fig. 3B).
The RBSDV P9-1 crystal structure revealed two molecules in the asymmetric unit, which form a noncrystallographic homodimer (27). In P9-1DC-arm, an identical dimeric arrangement is observed between two protomers belonging to neighboring asymmetric units (Fig. 3C, top). These polypeptide chains are related by a 2-fold symmetry axis, and the dimer can be constructed by means of the y, x, -z symmetry element of the P4 3 2 1 2 space group.
The P9-1DC-arm dimeric arrangement, supported by the SEC-SLS and DLS experiments described above ( Fig. 2C and D), is stabilized by an interface area of 693 Å 2 (4.9% of the total solvent-accessible surface per protomer) according to the PDBePISA server (33). In the dimeric assembly, the hairpins bF-bG and the loops bA-bB and bD-bE from both protomers protrude from the main body of the dimer in nearly the same direction (Fig. 3C, bottom). The P9-1DC-arm dimer shows a length of 8.5 nm in its largest dimension (Fig. 3C, bottom), which is consistent with the D h of ;8.4 nm estimated by the DLS measurements ( Fig. 2D; Table S1).
Cryo-EM analysis shows that P9-1 multimerizes as pentamers and hexamers of dimers with an internal pore. Given (i) our previous results in which more complex oligomeric structures were identified in the full-length protein in solution ( Fig. 2; Fig. S1 and S2) and (ii) the fact that RBSDV P9-1 forms an octamer where adjacent dimers are related by a 4-fold axis through their C-arms (27), we further performed single-particle cryo-EM studies on the full-length P9-1 protein.
The careful analysis of the recorded data clearly exposed doughnut-shaped (torus topology) pentamers of homodimers (10-mer D5 symmetry) and hexamers of homodimers (12-mer D6 symmetry), representing 80 to 85% and 15 to 20% of the particle populations, respectively (Fig. 5A). These data are consistent with the oligomerization state of ;10.4 protomers estimated by the SEC-SLS assays shown above (Fig. 2B).
The global resolutions for each species were 4.7 Å (decamer) and 6.8 Å (dodecamer) based on the "gold standard" criterion (Fourier shell correlation [FSC] = 0.143), with local resolutions ranging from 2.5 to 4.5 Å (decamer) and from 5.5 to 9.5 Å (dodecamer). We refer the reader to the Materials and Methods and Table 2 and Fig. S5 for details on data acquisition, data processing, and map statistics.
The P9-1DC-arm dimer crystallographic structure perfectly fit as a rigid body into the respective EM density maps ( Fig. 5B; Fig. S6). In addition, density protrusions corresponding to the C-arm regions (residues 314 to 337) were clearly distinguishable among the docked dimers on both quaternary assemblies ( Fig. 5A and B). Thus, the respective C-arms were traced and real space refined in both density maps along with the docked dimer crystallographic structures to obtain the complete atomic models ( Fig. 5B and C; Fig. S6), which showed very good refinement statistics and stereochemistry ( Table 2).
The full-length dimers are settled in a parallel mode related by the respective 5-fold and 6-fold rotational symmetry axes holding their C-arm protrusions as staplers, as previously reported for the octameric arrangement of the RBSDV P9-1 crystal structure (27). In this way, the C-arm of subunit A interacts with the neighboring (21) dimer, while the Carm of subunit B interacts with the other-side (11) dimer (Fig. 5B). The loops bA-bB (RNA binding site) and bD-bE (PEST motif), partially defined in the density maps as observed in the P9-1DC-arm crystallographic structure, protrude to the internal pore at the central part and the extremes (top and bottom) of the two oligomeric structures, respectively (Fig. 5C).
The nascent portion of the C-arm is sandwiched between the b-sheet 3 (hairpin bF-bG 1 b-strand H) from one adjacent subunit and the loop bA-bB along with the helix aV from the other, while the distal portion is partially embedded on the surface of the latter nearly aligned to the helix aIV (Fig. 5C). According to the PISA server (33), 17 residues (;70% of the C-arm total extension) are part of the intermolecular contacts with the adjoining dimer, mostly stabilized by hydrophobic forces. Structural comparison with the RBSDV P9-1 crystal structure revealed a dislocation of the C-arm trajectory with respect to both quaternary assemblies, where the hinge point is noticeable from Val316 in the nascent C-arm backbone (Fig. 5D). Interestingly, despite the high similarity found in the C-arm sequence of homologous proteins, changes are noted in this particular region, where a valine residue (Val316) is exclusively found in P9-1, while a threonine residue (Thr326 in RBSDV P9-1) is conserved in other closely related fijivirus proteins (Fig. S4).
Small-angle X-ray scattering (SAXS) analysis provides further evidence of the decameric and dodecameric states of P9-1. SAXS was used to gain further insights into the solution behavior of the P9-1 quaternary assemblies. A thorough analysis of the collected data revealed a small fraction contribution of larger aggregates to the final scattering curve. Fortunately, the fraction of larger aggregates was found to be minor, which allowed us to obtain useful insights.
The compact, globular nature of the P9-1 higher-order oligomers was confirmed by normalized Kratky analysis (Fig. 6A, inset), which is fully consistent with the previously described decamer and dodecamer structures. Analysis of the final scattering curve with OLIGOMER (34) revealed that P9-1 decamers and dodecamers were the predominant species in solution (fractions of 87% 6 1% and 13% 6 1%, respectively), with a small contribution by larger aggregates. The decamer proportion decreased to 75% 6 5% when a full pattern modeling was used (Fig. 6A). The two possible oligomers exhibit different features in the Porod region, which were useful to estimate the volume fraction of decamers and dodecamers, despite the Guinier region being partially affected by the presence of larger aggregates. The estimation of the MM from the simulated oligomer SAXS patterns (35) using an average density for large proteins (36) was 407 6 40 kDa for the decamer and 508 6 50 kDa for the dodecamer (Fig. 6B). These values were expected for these oligomers, as their theoretical MMs are 449.0 and 538.8 kDa, respectively. Thus, the simulated patterns from cryo-EM density maps were considered representative. The larger aggregates showed a fractal dimension of 2 (platelet like), which was also observed in other samples with larger aggregates (Fig. S7A), where the aggregation degree did not change with dilution (strong particle-particle interaction). There was an additional structural aspect to be considered, as the water-ion affinity may change inside the oligomer pore with respect to the outer protein surface. Thus, solution density inhomogeneity changed the scattering contrast and increased the estimation error of the volume fraction (Fig. S7B). The final scattering curve was also used for ab initio modeling (Fig. S7C). Distance distribution analysis revealed a maximal particle dimension (D max ) of 183 Å, and shape reconstruction with P2 imposed symmetry resulted in a low-resolution model consistent with the dimensions of the decamer and dodecamer observed.
In conclusion, the SAXS data were consistent with the above-mentioned biophysical and structural data and support the simultaneous presence of decameric and dodecameric quaternary P9-1 states in solution, with the former being the predominant species as indicated by cryo-EM.
P9-1 C-arm-mediated oligomerization into (do)decamers favors RNA binding. To explore whether the nucleic acid binding activity of P9-1 depends on the C-arm, increasing amounts of purified P9-1 and P9-1DC-arm were incubated with a 22-mer FIG 5 Legend (Continued) according to secondary structure elements as in Fig. 3A. The loops bA-bB and bD-bE, which protrude to the internal pore at the middle and the extremes of the structure, respectively, the helices aIV and aV, and the N and C termini are labeled. Curved dashed lines indicate the disordered regions in the reconstruction of the loops. For clarity purposes, the C-arms are highlighted by the solvent-accessible surface (transparent blue) in the background calculated by PyMOL. Scale bars are shown. Protomers marked with asterisks in C and D represent analogous views for clarity. (D) Structural comparison among individual full-length protomers of P9-1 decamer (10-mer, orange), P9-1 dodecamer (12-mer, blue), and RBSDV P9-1 (8-mer, green). A bar indicates the C-arm regions. The change noted in the C-arm trajectory (hinge point) among the three promoters is highlighted by a dashed rounded rectangle. The residues found at the dislocation (Val316 in P9-1 and Thr326 in RBSDV P9-1) are depicted as sticks and are colored according to their corresponding promoters.
Structural and Functional Characterization of Mal de Río Cuarto Virus Protein P9-1 mBio hexachloro-fluorescein (HEX)-labeled single-stranded DNA (ssDNA) probe and subjected to electrophoretic mobility shift assays (EMSA). The migration of the protein-ssDNA complexes was monitored by fluorescence detection of the probe, and protein migration was monitored by staining the native gel with Coomassie brilliant blue (Fig. 7A). As expected (16), P9-1 bound ssDNA in a concentration-dependent manner, showing a statistically significant increment in DNA binding between 1, 3, and 6 mM protein. Coomassie staining confirmed that the complexes shifted according to the migration of the multimeric assemblies of P9-1. P9-1DC-arm also bound ssDNA, and the protein-DNA complexes were less retarded, in agreement with the migration pattern of the dimers formed by this protein construct (Fig. 7A). Like P9-1, ssDNA binding by P9-1DC-arm was dependent on the protein concentration. As a negative control, bovine serum albumin (BSA) did not bind ssDNA. These results indicate that the binding of a 22-nucleotide (nt)-long ssDNA is independent of the presence of the C-arm (and thus P9-1 higher-order oligomeric states), consistent with the potential nucleic acid binding site residing within the loops bA-bB.
Because ssRNA binding activity is crucial for reovirus replication within viroplasms, we next performed competition assays by adding increasing amounts of unlabeled long ssRNA (an average of 250 nt long) to the P9-1-ssDNA and P9-1DC-arm-ssDNA complexes. Competition EMSAs showed that both proteins bind long ssRNA, but P9-1 binding is more efficient (Fig. 7B). Semiquantitative analysis of the labeled ssDNA band patterns revealed a 10-fold decrease in the P9-1-ssDNA complex band at 100 mM ssRNA, as opposed to a lower 1.5-fold decrease for the P9-1DC-arm-ssDNA complex (Fig. 7B). Thus, the presence of the C-arm strongly favors RNA binding, probably as a result of P9-1 oligomerization into (do)decamers. In addition, P9-1 presented a marked retarded migration pattern at increasing ssRNA concentrations (Fig. 7B), suggesting the binding of multiple P9-1 (do)decamers. To assess if this behavior is dependent on ssRNA length, 6.5 mM P9-1 was incubated with 250 mM 30-nt Cy5-labeled ssRNA. Under these conditions, the retarded migration pattern was not observed (Fig. S8), indicating that this effect is dependent on long RNA molecules. ATPase activity is stimulated by the binding of RNA to P9-1 (do)decamers. Because it is known that P9-1 catalyzes ATP hydrolysis (16), we quantitatively determined whether deletion of the C-arm affects ATPase activity as well as if the binding of ssRNA to P9-1 and P9-1DC-arm has an effect on such catalytic activity (Table 3; Fig. S9). At a protein concentration similar to the one used in the nucleic acid-binding assays (6.5 mM), both proteins presented similar enzymatic activity values, whereas nonenzymatic hydrolysis was negligible. Interestingly, in the presence of ssRNA (500 mM), the P9-1 ATPase activity increased five times, while no detectable effect was observed with P9-1DC-arm. These results are indicative of an interaction between the RNA and ATP binding sites, which results in an RNA-dependent ATPase activity enhancement in P9-1 (do)decamers that is dependent on the presence of the C-arm.
In silico simulations are compatible with a strong binding of phosphate to the P9-1 pore and the C-arm, which is enhanced in the decameric form. To further characterize the RNA binding and ATPase activities, we computed classical molecular interaction potentials on the P9-1 dominant quaternary conformations using spherical probes mimicking the phosphate groups found in RNA and ATP. 21858138, respectively). As defined above, the missing regions are mostly composed of the flexible loops bA-bB and bD-bE that protrude to the internal pore of the (do)decameric assemblies. Interestingly, as a result of the loop orientations, the pore is more occluded in the decamer than in the dodecamer (Fig. 8A, middle and right). We found a marked ability of P9-1 to bind phosphate moieties in the three multimeric states, with a strong preference for the decamer (Fig. 8B). The most probable binding sites at the 25 kcal/mol isosurface (enthalpic interaction energy) were located inside the pore involving part of the loops bA-bB. Remarkably, a significant binding probability was found on top and around the C-arm region, which was again enhanced in the decameric form (Fig. 8C).

DISCUSSION
Biochemical and structural analyses of reoviral viroplasm proteins are beginning to unravel functional aspects of viroplasm maturation and dynamics. Within the Reovirales order, the major viroplasm proteins from rotavirus (NSP2), phytoreovirus RGDV (Pns9), and the fijiviruses RBSDV (P9-1) and SRBSDV (P9-1) present ring-or doughnut-shaped octameric structures (21,(26)(27)(28). In this study, we showed that MRCV P9-1 gives rise to pentamers and hexamers of dimers (10-mers and 12-mers, respectively), which resemble the overall quaternary structure folding previously reported. Although these arrangements have not been previously described at the atomic level in other related proteins, the structural characterization of the BTV NS2 protein by negative-stain EM revealed A two-way ANOVA followed by Tukey's multiple-comparison test was performed. In the right graph from A, different letters denote statistically different values, while in the right graph from B, statistically different values between proteins at each ssRNA concentration are denoted by asterisks (P , 0.05).
Structural and Functional Characterization of Mal de Río Cuarto Virus Protein P9-1 mBio similar ring-like assemblies that could correspond to decamers or dodecamers (25), suggesting that these higher-order oligomeric structures may also occur in other reovirids. The P9-1 C-arm does not affect dimer assembly but instead is critical for the formation of the decameric and dodecameric quaternary structures. Our findings led us to propose a model for the arrangement of the oligomers. Initially, two monomers of P9-1 would interact to form a dimeric assembly across a surface of 20 residues stabilized mainly by hydrogen bonds and hydrophobic interactions. The P9-1DC-arm and P9-1 dimer interfaces, almost identical to those described in full-length RBSDV P9-1 (27) and SRBSDV P9-1 (28), support that the dimer assembly is independent of the C-arm and suggests it would be conserved across major viroplasm proteins from plant reoviruses. Following the dimer assembly, five or six dimers would interact through their C-arms to give rise to the (do)decameric structures, with no intermediate oligomeric states of lower molecular masses like the dimeric and tetrameric species observed in RBSDV P9-1 (27,29).
The versatility of the P9-1 structure, allowing either five or six dimers to self-assemble, denotes the flexible nature of the P9-1 C-arm. This is also evident when comparing the protomers of P9-1 (do)decamers with those of the octameric RBSDV P9-1. Identifying the possible structural elements associated with the changes in the C-arm trajectory is not straightforward, as RBSDV P9-1 is the only available fijivirus P9-1 structural model containing the C-arm. One possible scenario may involve the presence of a valine residue in the highly conserved C-arm region, which is present in MRCV. Forthcoming full-length structural models of homologous proteins will help determine whether the C-arm flexibility is an exclusive structural feature of P9-1 or if it is a feature shared with other fijivirus homologous proteins.
The structural flexibility of the C-arm along with the subtle protein contacts found around it in the 10-mer and 12-mer oligomers may provide a fine-tuned dynamic assembly/disassembly mechanism that could be crucial for P9-1 function. Indeed, a previous study has reported that NSP2 adopts a slightly looser octameric conformation, which partially dissociates into tetramers in the presence of magnesium (37). Moreover, NSP2 and NS2 both undergo conformational changes after binding to a nonhydrolyzable ATP analog (25,37), suggesting that these proteins have dynamic structures.
Our experimental and theoretical findings agree with previous studies, which reported that RBSDV P9-1 preferentially binds ssRNA in a multimeric conformation, where the major RNA binding site is located at the inner pore involving the loops bA-bB (27,29). The conserved nascent region of the C-arm was predicted to be a second RNA binding site in RBSDV P9-1 (Fig. S4). However, this could not be experimentally assessed because of the critical role of the C-arm in multimerization, which affects the RNA binding affinity (29). Due to the proximity between the loops bA-bB and the nascent region of the C-arms within the P9-1 (do)decameric structures, we suggest that the RNA interaction may involve both regions.
P9-1DC-arm showed similar ATPase activity to the full-length protein, revealing that the C-arm is not part of the ATP catalytic site. However, because the ATPase activity is not stimulated by RNA in this protein construct, a C-arm-mediated allosteric communication between the ATP catalytic site and the RNA binding site should exist. In other words, because the C-arm plays a crucial role in P9-1 oligomerization, the binding to RNA would allosterically modulate the ATP catalytic site exclusively when the protein is assembled into (do)decamers. The question remains whether the two quaternary arrangements described here differently affect the functional communication between the ATP and RNA binding sites. In this regard, the crowding effect within the pore that takes place in the decamer would promote the interaction between both sites, as additional and stronger phosphate binding sites were predicted by in silico simulations. This finding gives a potential functional advantage for the decamer over the dodecamer, which may be endorsed by the cryo-EM and SAXS experimental data, where the decameric assembly was the most represented species.
Reovirids packaging into nascent virions involves a precise order of intersegment interactions between their 1RNA genome segments and an RNA chaperone activity (6,31). The multimeric structures of P9-1, resembling the ring-like shape of hexameric helicases (38), and its ATPase and RNA binding activities suggest that this protein participates in the equimolar copackaging of the viral genome. In fact, a previous study showed that the orthoreovirus protein s NS is a viroplasm component that acts as an RNA chaperone facilitating RNA-RNA interactions during genome packaging (39). Similarly, rotavirus NSP2 viroplasm protein has been shown to be an RNA nonspecific chaperone that binds viral ssRNAs and promotes stable intersegment contacts that are required for viral genome packaging (40). Future studies will define the precise role of P9-1 in the virus infection cycle.
It has recently been shown that the formation of rotavirus viroplasms occurs via liquid-liquid phase separation (LLPS) of viroplasm-forming proteins (41). This phenomenon is favored in proteins harboring IDRs and RNA binding properties (42,43). Interestingly, P9-1 contains two IDRs, and the structural models described herein confirm that they are located at the flexible loops bA-bB and bD-bE. In addition, we provided evidence that $250-nt-long ssRNA would allow the binding of multiple P9-1 (do) decamers. These results, together with the dynamics of the formation of VLSs upon P9-1 expression in insect and plant cells, support the hypothesis that the MRCV viroplasm would also be formed via LLPS.
Studies have previously shown that in rotaviruses and avian reoviruses, viroplasm formation requires a functional proteasome (44)(45)(46). P9-1 was shown to contain a PEST motif whose removal favors protein accumulation (18). PEST motifs can be activated by different mechanisms, such as light, ligand binding, or phosphorylation (20). The location of P9-1 PEST sequences within the flexible loops bD-bE, exposed toward the top and bottom of the inner pore of the (do)decameric structures, suggests that regulation of PEST sequences could affect P9-1 conformational stability, accelerating proteasomal degradation (47).
In conclusion, the work presented here describes structural conformations previously unidentified for a major viroplasm protein and provides evidence at the molecular level that it may simultaneously adopt two distinct quaternary assemblies. In particular, these findings illustrate the structural versatility of P9-1 and raise the question of whether the distinct homooligomeric structures have different biological properties during the virus infection cycle. Furthermore, this work reveals an allosteric communication between ATP and RNA binding sites, deciphering a potential functional feature in reovirids viroplasm proteins.

MATERIALS AND METHODS
Cloning of expression plasmids. The reported pRSET P9-1 construct containing the P9-1 coding sequence (UniProt accession number D9U542; 337 residues, 39 kDa) in frame with a sequence encoding Structural and Functional Characterization of Mal de Río Cuarto Virus Protein P9-1 mBio a 52-residue N-terminal tag (6ÂHis/Xpress/enterokinase cleavage recognition sequence, EK) was used for P9-1 recombinant expression and purification in bacteria (16). The previously reported P9-1 biochemical characterization, including homomultimerization, ATPase, and ssRNA-binding activities were performed with this recombinant construct, which encodes a 44.9-kDa protein (16). Next, this construct served as a template for a PCR designed to clone the P9-1DC-arm (residues 1 to 313 of P9-1, lacking 24 residues at the C terminus) with an N-terminal 6ÂHis tag into the pET24a vector. The previously described pCR8/GW/TOPO (Invitrogen, USA) entry vectors containing the P9-1 and P9-1DC-arm (18) coding sequences were used for recombination with the LR Clonase II enzyme mix (Invitrogen, USA) according to the manufacturer's instructions. For live imaging in transfected rice protoplasts, the pUC57-43 vector was used (48), whereas for live imaging in transfected Sf9 insect cells, the pIB-GW destination vector (17) was used. The resulting constructs express P9-1 or the P9-1DC-arm construct (lacking the 24 C-terminal residues) fused to the green fluorescent protein at the N terminus (GFP: P9-1 or GFP:P9-1DC-arm).
Additionally, a recombinant baculovirus for P9-1 expression in infected Sf9 cells was obtained using the Bac-to-Bac system (Invitrogen, USA). The coding sequence of P9-1 was excised from pRSET and directionally cloned into pFastBac. The resulting recombinant protein (P9-1 Sf9) has an N-terminal tag of 44 residues containing a 6ÂHis tag and a molecular weight of 44.12 kDa.
Fluorescence imaging was performed in a Leica TCS-SP5 (Leica Microsystems GmbH, Germany) spectral laser confocal microscope using a 63Â objective (HCX PL Apo CS 63.0 1.20 water UV). The 488-nm line of the argon laser was used for GFP excitation, and the fluorescence emission was detected with channel settings of 498 to 540 nm for GFP.
Protein production and purification. The P9-1 and P9-1DC-arm constructs for bacterial expression were transformed into Escherichia coli BL21 cells grown in Terrific Broth culture medium supplemented with 0.1% glucose and 50 mg/mL ampicillin and induced for expression with 1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) at 28°C overnight. Cells were harvested by centrifugation for 15 min at 5,000 Â g and 4°C and resuspended in lysis buffer (20 mM sodium phosphate, 500 mM sodium chloride, 20 mM imidazole, 1 mM phenylmethylsulfonyl fluoride [PMSF], 0.05% Triton X-100, and 100 mg/mL lysozyme, pH 7.4) using 10 mL of lysis buffer per 100 mL of cell culture. Soluble proteins were obtained by sonication with 3 pulses of 30 s each in an ice bath using a Vibra-Cell ultrasonic liquid processor (Sonics & Materials, Inc., USA) and centrifugation at 12,000 Â g for 15 min at 4°C. A second extraction was performed by resuspending the remaining pellet in 5 mL of lysis buffer per 100 mL of cell culture.
The recombinant baculovirus for P9-1 expression in eukaryotic Sf9 cells was purified with a ZR BAC DNA miniprep kit (Zymo Research, USA) and transfected in Sf9 cells by using Cellfectin II reagent (Invitrogen, USA), following the manufacturer's instructions. Baculovirus stocks were obtained by infecting Sf9 cells (grown in 75-cm 2 flasks) at a multiplicity of infection (MOI) of 0.05 and harvesting at 4 days postinfection (4 dpi). Viral titers were determined by the endpoint dilution method. For recombinant protein expression, 1.3 Â 10 8 Sf9 cells grown in suspension cultures in shake flasks were infected at an MOI of 2, and at 4 dpi, cells were harvested by centrifugation at 500 Â g for 10 min. Cells were then resuspended in 30 mL of lysis buffer (50 mM Tris-HCl, 150 mM sodium chloride, 1 mM EDTA, 0.01% Triton X-100, 0.5% NP-40, and 1 mM PMSF, pH 7.5), and proteins were extracted by centrifugation at 12,000 Â g for 20 min at 4°C.
Protein extracts were subjected to IMAC by incubation with 2 mL of nickel-nitriloacetic acid (Ni-NTA) resin (Qiagen, Germany) per 50 mL of extract for 3 h at 4°C with gentle agitation. After incubation, the resin was loaded on an empty column and washed with lysis buffer, and the bound protein was eluted with 20 mM sodium phosphate, 500 mM sodium chloride, and 500 mM imidazole (pH 7.4). Buffer exchange and protein sample concentration were performed using 10-kDa molecular weight cutoff (MWCO) Vivaspin Turbo centricons (Sartorius, Germany). After IMAC, proteins were further purified by SEC using a Superdex 200 column (GE Healthcare, USA) in running buffer (10 mM Tris-HCl and 25 mM sodium chloride, pH 7.6) at a flow rate of 1.3 mL/min, followed by another concentration step with 10-kDa MWCO Vivaspin Turbo centricons. Protein quantification was assessed using a spectrophotometer (NanoDrop 1000, Thermo Fisher Scientific, USA).
SEC-SLS measurements. The average MMs of P9-1 and P9-1DC-arm in solution were determined on a Precision Detectors PD2010 90°light scattering instrument tandemly connected to high-performance liquid chromatography and an LKB 2142 differential refractometer. The chromatographic runs were performed in a Superdex 200 GL 10/300 column (GE Healthcare) with a buffer containing 10 mM Tris-HCl and 25 mM sodium chloride (pH 7.6) at a flow rate of 0.4 mL/min. Elution was monitored by measuring the SLS signal at 90°and its refractive index (RI). The masses of the injected samples were 150 mg for P9-1 and 300 mg for P9-1DC-arm. The MM of each sample was calculated relating its SLS and RI signals and comparing this value with the one obtained for bovine serum albumin (MM: 66.5 kDa) as a standard. Data were analyzed with the Discovery32 software supplied by Precision Detectors. The average MM value corresponded to the central 10% of the peak.
DLS measurements. The size distribution and hydrodynamic diameter measurements were performed at 25°C with a Zetasizer Nano-S DLS apparatus (Malvern Instruments Ltd., UK) using a low-volume quartz cuvette. Protein samples were diluted to ;2 mg/mL in 25 mM Tris-HCl and 100 mM sodium chloride (pH 7.5). For each sample, 7 to 10 runs 10 s in length were performed. Size distributions and hydrodynamic diameters were calculated using the multiple narrow distribution analysis models of the DTS v.7.11 software (Malvern Instruments Ltd., UK).
Analytical SEC. Analytical SEC was performed using an ENrich 650 10/30 column (Bio-Rad, USA) preequilibrated in a running buffer (25 mM Tris-HCl and 100 mM sodium chloride, pH 8.0). Bio-Rad gel filtration calibration standard composed of bovine thyroglobulin (670 kDa), bovine g -globulin (158 kDa), chicken ovalbumin (44 kDa), horse myoglobin (17 kDa), and vitamin B 12 (1.35 kDa) was used as MM standard, although the elution volume of the latter was excluded from the analysis. The P9-1 protein sample (500 mL) was injected at 1 mg/mL and eluted at a flow rate of 0.5 mL/min. Calibration of the column was performed using the Bio-Rad MM standard under the same conditions, and the apparent MM of P9-1 was determined according to reference 50. The partition coefficient (K) was calculated as where V x is the elution volume of each standard protein, V 0 is the void volume, and V e is the end volume of the column. Estimation of the experimental D h of P9-1 was based on the elution volumes and the D h of the standard proteins, given by the relationship 1,000 Crystallization, X-ray data collection, and structure resolution of P9-1DC-arm. Initial crystallization conditions for P9-1DC-arm were screened at room temperature on 96-well sitting-drop Greiner 609120 plates using a Digilab Honeybee963 robot (Marlborough, USA) and commercial kits from Jena Bioscience (Jena, Germany) and Hampton Research (Aliso Viejo, USA) at a protein concentration of 15 mg/mL. Optimized crystals were then grown in 24-well hanging-drop Hampton Research VDX plates with a precipitation solution consisting of 13% (wt/vol) polyethylene glycol 8000 (PEG 8000) and 0.2 M calcium acetate, reaching a maximum size of 0.4 Â 0.1 Â 0.1 mm 3 . Several detergents and additives (Hampton Research) were tested around this condition, but none of them significantly improved the crystal size and/or diffraction quality. Crystals were cryoprotected in mother liquor supplemented with 22% (wt/vol) PEG 400 and flash-cooled in liquid nitrogen using Hampton Research loops.
X-ray diffraction data sets were collected at 100 K on several crystals at the PROXIMA-2A beamline at Synchrotron SOLEIL (France) using an EIGER X 9M detector (Dectris, Baden, Switzerland) and the MXCuBE application (51). The best crystal diffracted to a resolution of 3.47 Å (Table 1). Data sets were indexed, integrated, and scaled with XDS (52), leaving 5% of the reflections apart for cross-validation. The P9-1DC-arm structure was solved by molecular replacement with Phaser (53) using the coordinates of RBSDV P9-1 as a search model (PDB code: 3VJJ) (27). Refinement and manual model building were then performed with the programs Buster (54) and Coot (55), respectively. Due to the low resolution, on the initial refinement cycles, specific reference model restraints using RBSDV P9-1 as a template along with an automatic setting of the relative weight between geometry and X-ray terms were applied to ensure the correctness of the model. Intermediate refinement steps performed on the PDB_REDO server (56) were critical for structure model optimizations. The final model was validated with MolProbity (32) as well as with the validation module implemented in Coot (55). Table 1 summarizes the statistics generated at these steps.
Cryo-EM data acquisition of full-length P9-1, data processing, model building, refinement, and validation. High-quality recombinant P9-1 protein samples were suspended in buffer (10 mM Tris-HCl and 25 mM sodium chloride, pH 7.6) at 17 mg/mL and kept on ice before cryo-grid preparation. Several serial dilutions were prepared, and 3 mL of each sample was loaded on Quantifoil R2/2Cu/Rh 300 holeycarbon-supported grids (Quantifoil Micro Tools GmbH, Jena, Germany). Initial cryo-EM sample preparations showed clear protein aggregation, which was reverted by omitting the glow-discharge step on the grids. The samples were incubated with the grids for 1 min, blotted by filter papers, and plunge-frozen into liquid ethane cooled by liquid nitrogen using a Leica EM CPC manual plunger. The vitrified grids were stored in liquid nitrogen for later use. The best grids were obtained at 1/10 (vol/vol), where homogenous and well-spread individual particles were clearly identified. Data acquisition was performed using a Talos Arctica microscope (Thermo Fisher) operated at 200 kV with an FEI Falcon III direct detector at Centro Nacional de Biotecnología (CNB; Spain) for 1 day per grid using a nominal magnification of 120,000, corresponding to a calibrated pixel size of 0.855 Å per pixel and a defocus range of 20.8 to 23.8 mm. A total number of 669 micrograph movies were recorded under low-dose conditions and fractionated into 60 frames each with a dose of 0.5 e -/Å 2 per frame. All data processing was executed using Scipion (59), a software framework integrating several 3DEM software packages, as detailed below.
Micrographs were aligned for motion correction purposes and dose weighted with MotionCor2 (60). Determination of the Contrast Transfer Function (CTF), beam-induced movement, defocus values, astigmatism, and micrograph resolution estimation were performed using Ctffind4 (61). The final images were carefully examined for further image processing considering the particle distribution, the resolution, and the quality of Thon ring fitting. An initial template-free particle picking was performed (first manually and then automatically) using Xmipp 3.0 (62). The preliminary set of picked single particles (202,824 particles) was first subjected to an initial two-dimensional (2D) classification, resulting in 163,663 particles with 5-fold symmetry and 30,526 particles with 6-fold symmetry (Fig. S5). Next, single 2D class averages were used as references using Eman2 (63) for getting two preliminary ab initio volumes with D5 and D6 symmetries, which were used as a reference for 3D classification and refinement. After several rounds of refinement, two clearly different and well-populated 3D classes, decamer (with imposed D5 symmetry) and dodecamer (with imposed D6 symmetry), were found. The particles were then further extensively 3D classified using Relion-3 (64), resulting in a major population of 99,682 (decamer) and 22,510 (dodecamer) particles. Reconstructions of the final maps were sharpened by dividing the maps by the modulation transfer function of the detector and by applying a negative B-factor using Relion-3 (64). Local resolutions of the maps were calculated using ResMap (65). The data processing workflow is described in Fig. S5, and the data collection and reconstruction statistics are shown in Table 2.
The dimeric P9-1DC-arm crystal structure was fitted as a rigid body into the respective EM density maps using UCSF ChimeraX (66). Later, the C-arm region (residues 314 to 337) was traced using the RBSDV P9-1 crystal structure (PDB code: 3VJJ) as a reference and manually adjusted using Coot (55). The docked atomic coordinates of the respective 3D models were refined into the locally filtered maps using phenix.real_space_refine with secondary structure restraints calculated in Phenix 1.18.2_3874 (67). The validation of the models was performed using the MolProbity software (32). Model building and refinement statistics are shown in Table 2.
SAXS analysis. P9-1 SAXS measurements were performed at the DO1B-SAXS1 beamline of the Brazilian Synchrotron Light Laboratory (LNLS, Brazil) with an incidence wavelength (l) of 1.54 Å. The scattering intensity distributions as a function of the momentum transfer q were obtained in the q range between 0.013 and 0.48 Å 21 with q = 2p sin(u )/l, where 2u is the scattering angle. The SAXS patterns were recorded with exposure times of 20 s per frame for 10 min. A Pilatus 300K detector was used with an 883-mm sample detector distance. One-dimensional curves were obtained by integration of the 2D data using the program FIT-2D (68). Liquid samples were injected into the beamline vacuum-tight temperature-controlled X-ray cell for liquids. The P9-1 fractions obtained from exclusion chromatography were diluted from 5 to 0.5 mg/mL, and no change in SAXS patterns with dilution was observed within this range of concentrations. Simulated patterns of the individual protein oligomers were done using pseudoatom approximation obtained from low-resolution refinement of cryo-EM experiments with a combination of the Scipion platform (59, 69) and ATSAS 2.1 package (70). Simulated patterns were used for data interpretation using least square procedures. Because a small proportion of larger aggregates were observed after purification, a fractal aggregate was included (71). Also, a Gaussian chain form factor (71) was used as a background function to account for flexible parts of the proteins. The OLIGOMER package (34) was also tested to estimate each oligomer volume fraction. Ab initio modeling was performed with the DENSS package (72) with imposed P2 symmetry.
EMSA. Various amounts of purified P9-1, P9-1DC-arm, or BSA (as a negative control) were incubated with a 22-nt HEX-DNA oligonucleotide probe (5 mM; 59-HEX-GACCTCGCTCTCTGTTTCTCAT-39) in buffer (10 mM Tris-HCl, 50 mM potassium chloride, 0.5 mM EDTA, 10% glycerol, 1 mM dithiothreitol [DTT], pH 7.5) (16). Different concentrations of poly(A) (polyA; Midland Certified Reagent Company, USA) were used for ssRNA competition experiments. According to the supplier, it consisted mostly of poly(A) polymers of an average of 250 nucleotides in length. Reactions were held for 30 min at room temperature in a total volume of 20 mL and subjected to 6.5% native PAGE run in a cold room (4°C). Migration of the labeled probe was detected in a XX6 G-box imaging system (Syngene, USA). Three independent experiments were performed (n = 3). The fluorescence intensity of the ssDNA probe in complex with the proteins was quantified with ImageJ software (73). Because the complexes migrated differently for P9-1 and P9-1DC-arm, an extensive rectangular area was selected to comprise the intensity of the probe in complex with the two proteins for every lane in succession (indicated in Fig. 7A, left, as "Protein-ssDNA complexes"). The statistical significance of the signal was calculated using a two-way analysis of variance (ANOVA) followed by Tukey's multiple-comparison test with GraphPad Prism version 8.0.0 for Windows (GraphPad software, San Diego, CA, USA; www.graphpad.com).
The binding experiments with short RNA were performed using a Cys5-labeled 30-nt oligonucleotide (59-Cy5-CAUCAUGCAGGACAGUCGGAUCGCAGUCAG-39) that was incubated with 6.5 mM protein and subjected to native PAGE under the conditions described above.
ATPase activity measurements. The initial rate of ATP hydrolysis for P9-1 and P9-1DC-arm was obtained from the slope of the time course of inorganic phosphate release. Reactions were performed at 25°C in 25 mM Tris-HCl, 100 mM sodium chloride, 0.5 mM EDTA, and 4.4 mM magnesium chloride (pH 7.7). All reactions were initiated with the addition of 2.5 mM ATP after protein preincubation for 10 min at 25°C in reaction medium and a protein concentration of 6.5 mM. All reactions were stopped by the addition of ammonium heptamolybdate solution in an acidic medium, and the amount of inorganic phosphate was quantified spectrophotometrically according to the Baginski method (74) with modifications (75). The absorbance was measured in a Jasco V-550 spectrophotometer. When present in the reaction medium, 500 mM poly(A) ssRNA ( $250 nt) was added before the 10-min protein preincubation. Five different reaction times were used to determine the velocity and, to ensure initial rate conditions, the hydrolysis never exceeded 5% of the starting concentration of ATP. The spontaneous hydrolysis of ATP was followed under identical conditions without protein and was negligible under these conditions.
In silico reconstructions of the missing structural regions and MD simulations. The amino acid sequence of the full P9-1 dimer (674 residues) was aligned to the P9-1DC-arm crystal structure, and a 3D model was created with MODELLER (76) to rebuild the regions not defined in the electron density map and the C-arm residues (314 to 337). Protonation states at pH 7.5 were assigned by the PDB2PQR server (77). The Cartesian coordinates of all residues present in the crystal structure were fixed in space as found experimentally by means of a harmonic potential of 10 kcal/mol/Å 2 ; the remaining residues were allowed to move freely during all the following MD steps (see Fig. 8A, left, and Movie S1 available through Figshare at https://figshare.com/articles/media/mBIO-Llauger-et-al-2023-movieS1/21842169). The model was minimized in implicit solvent, neutralized with 16 Na 1 ions, solvated with explicit waters and 0.15 M NaCl, and minimized in solution. It was then thermalized to 299 K at constant mass, volume and temperature (NVT) and simulated during 4 ms using MD at constant mass, pressure and temperature (NPT) (P = 1 bar).
The minimized model of the full-length dimer (D1) was aligned to each of the dimers in decamer (D5) and dodecamer (D6) to reconstruct the whole structures; residues from the C-arm region were replaced by those experimentally determined. The Cartesian coordinates of all residues experimentally defined were fixed in space using a harmonic potential of 10 kcal/mol/Å 2 ; the remaining residues were allowed to move freely during all the following MD steps (see Fig. 8A, middle and right, and Movies S2 and S3 available through Figshare at https://figshare.com/articles/media/mBIO-Llauger-et-al-2023-movieS2/21842298 and https://figshare.com/articles/media/mBIO-Llauger-et-al-2023-movieS3/21858138, respectively). After minimization of the model in vacuo, some residues at the core of the structures and the C-arms were manually shifted to deinterlace regions from neighboring dimers, and a new minimization was run, removing the restraint over those residues. The models were then neutralized with 80 Na 1 ions (decamer) and 96 Na 1 ions (dodecamer), solvated with explicit waters and 0.15 M NaCl, and minimized in solution. Next, the systems were thermalized to 299 K at NVT and simulated during 3 ms (decamer) or 2 ms (dodecamer) by means of MD at NPT (P = 1 bar).
To treat every protein, we used the ff19SB force field (78), and the entire system was surrounded by a truncated octahedral box of TIP3P water molecules (79), applying Dang's parameters on ions (80). All systems were simulated using the Langevin algorithm to control the temperature and the pressure, with a coupling constant of 5 ps. SHAKE was used to keep all bonds involving hydrogen at their equilibrium values, which allowed us to use a 2-fs step for the integration of Newton's equations of motion. Longrange electrostatic interactions were accounted for by using the particle mesh Ewald method with standard defaults. All simulations were performed using the PMEMD CUDA code module of AMBER18 and analyzed with CPPTRAJ (81).
Molecular interaction potentials. The linear Poisson-Boltzmann equation (PBE) (without considering dielectric self-interaction), as implemented in CMIP (82), was used to compute free molecular interaction potentials using phosphate groups as probes. Experimentally determined structures rebuilt and simulated by means of MD simulations were used as initial structures for the protein complexes. Representative structures were chosen from the two most populated clusters (based on the RMSD fluctuations after convergence); that is, the two most prevalent conformations observed during MD simulations. The ionic strength was set to 0.15 M, and the reaction field dielectric constants for proteins and water were set to 4.0 (83) and 79.8, respectively. The van der Waals radii were taken from the ff99SB force field (78).
Statistical analysis. Statistical analyses were performed using GraphPad Prism version 8.0.0 for Windows (GraphPad software, San Diego, CA, USA; www.graphpad.com). Differences in values between study groups were assessed by analysis of variance (ANOVA) and Tukey's multiple-comparison test, and P values of ,0.05 were considered statistically significant.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. Authors declare that they have no competing interests.