Structural insights into the multifunctionality of rabies virus P3 protein

Significance

The multifunctional rabies virus P protein has diverse roles in replication and subverting host-cell processes, including evasion of innate immune responses. This functional diversity involves several P protein isoforms that have distinct phenotypes through interactions with various cellular proteins and structures. Here, we combine biophysical methods and cell-based assays to analyze the P3 isoform from pathogenic and attenuated viruses finding that, while some functions can be ascribed to specific domains, including intrinsically disordered regions, other functions are dependent on the three-dimensional arrangement and intramolecular interactions of these domains. This shows how complex molecular interactions can expand the functionality of proteins beyond their complement of individual functional modules, identifying a critical mechanism for viruses with a limited number of proteins.

Abstract

Viruses form extensive interfaces with host proteins to modulate the biology of the infected cell, frequently via multifunctional viral proteins. These proteins are conventionally considered as assemblies of independent functional modules, where the presence or absence of modules determines the overall composite phenotype. However, this model cannot account for functions observed in specific viral proteins. For example, rabies virus (RABV) P3 protein is a truncated form of the pathogenicity factor P protein, but displays a unique phenotype with functions not seen in longer isoforms, indicating that changes beyond the simple complement of functional modules define the functions of P3. Here, we report structural and cellular analyses of P3 derived from the pathogenic RABV strain Nishigahara (Nish) and an attenuated derivative strain (Ni-CE). We identify a network of intraprotomer interactions involving the globular C-terminal domain and intrinsically disordered regions (IDRs) of the N-terminal region that characterize the fully functional Nish P3 to fluctuate between open and closed states, whereas the defective Ni-CE P3 is predominantly open. This conformational difference appears to be due to the single mutation N226H in Ni-CE P3. We find that Nish P3, but not Ni-CE or N226H P3, undergoes liquid–liquid phase separation and this property correlates with the capacity of P3 to interact with different cellular membrane-less organelles, including those associated with immune evasion and pathogenesis. Our analyses propose that discrete functions of a critical multifunctional viral protein depend on the conformational arrangements of distant individual domains and IDRs, in addition to their independent functions.

RNA viruses have limited genomic space but can mediate the basic replication cycle as well as subvert the biology of infected cells, including antiviral immune responses, to create an environment conducive to replication. Central to this are multifunctional viral proteins, which can interface with multiple cellular compartments and proteins to regulate diverse cellular functions (1–4). The rabies virus (RABV, genus Lyssavirus, family Rhabdovirus) phosphoprotein (P protein) is an archetype of multifunctional viral proteins. RABV has only 5 genes (N, P, M, G, and L), but increases its coding capacity by expressing five P protein isoforms from the P gene. These include the full-length protein, P1 (residues 1 to 297), which is essential for replication as a cofactor for the viral RNA polymerase L protein, and the N-terminally truncated P3 protein (residues 53 to 297), which is generated by translation from an internal in-frame start codon and has a highly divergent phenotype, lacking L-binding/polymerase cofactor function (Fig. 1A). While P1 localizes primarily to the cytoplasm due to a strong N-terminal nuclear export sequence (N-NES, residues 49 to 58), P3 can localize to the cytoplasm, nucleus, and several liquid–liquid phase separation (LLPS)-dependent membrane-less organelles (MLOs) including microtubule (MT) bundles and, within the nucleus, nucleoli and nuclear bodies containing P3 and PML protein (5–9). The distinct localization of P3 is due to truncation of the N-NES, active nuclear import via a noncanonical N-terminal nuclear localization sequence (N-NLS, requiring sequence from residue 53, extending beyond residue 151, such that sequence encompassing residues 53 to 174 comprises activity comparable to full-length protein) (10, 11), and a conformational C-terminal NLS (C-NLS, residues 211 to 214, 260). P3 also contains NES activity in the C-terminal domain (P_CTD) due to the presence of a C-NES (residues 223 to 232) (11–13), such that nucleocytoplasmic localization depends on a balance of NLS/NES activity (7). These differences in the localization of P1 and P3 appear to enable multiple mechanisms of host-cell subversion, including in immune evasion, through interactions with cellular proteins of the innate immune system (8). The differing phenotype of P1 and P3, and the multifunctional nature of P proteins and similar proteins of other viruses and of cellular life, is typically attributed to a simple multimodular mosaic arrangement. Within this multimodular structure, independent sequences are considered to mediate different functions, and addition or removal of sequences produces a composite phenotype. However, several cell-based studies of P3 suggest more intricate mechanisms, potentially involving nonsequential conformational organization of functional sequences and domains, resulting in P3 gaining functions, such as binding to MTs and nuclear import receptors, which are not evident in longer isoforms (10, 11). Importantly, these functions have been implicated in lethal disease caused by RABV (6–8, 14). However, due in part to the complex molecular nature of P3, that includes several intrinsically disordered regions (IDRs) and globular domains (Fig. 1A), the broad structural organization of P3 and its significance in multifunctionality and pathogenesis are unknown.

Fig. 1.

Nish P3 and Ni-CE P3 show long-range conformational differences. (A) Schematic of domain structure of Nishigahara P3 (Nish P3) and positions of mutations (L56P, L58P, L66P, F81P, N226H) of P3 from the attenuated Ni-CE strain. IDR1 and IDR2 are regions of intrinsic disorder; DD and CTD are the well-structured dimerization and C-terminal domains, respectively. (B and C) Small-angle X-ray scattering data (B) and p(r) curves (C) for Nish P3 (black, brown), Ni-CE P3 (red, orange), and N226H P3 (blue, cyan). The p(r) curves are consistent with a smaller radius of gyration for Nish P3 (SI Appendix, Table S2). (D) Kratky plot for Nish P3 (black), Ni-CE P3 (red), and N226H (blue). (E–G) EOM distribution along with the models proposed for Nish P3 (E), Ni-CE P3 (F), and N226H P3 (G) showing that Nish P3 fluctuates between open and closed states, whereas Ni-CE P3 and N226H P3 show a predominant open state.

Here, we sought to determine the molecular differences in P3 proteins derived from a pathogenic RABV strain Nishigahara (Nish) and an attenuated derivative strain (Ni-CE) (6, 15), in which the respective P genes are determinants of the pathogenicity. The P3 proteins differ in immune evasion function, nuclear localization, and MT association, indicative of key roles in pathogenesis (6–8, 14), but the mechanisms underlying these differences are poorly understood, as the substitutions in Ni-CE P3 do not correspond to defined residues of P3 trafficking or localization motifs (7). We performed an analysis of the conformational ensembles adopted by P3, and structurally and functionally characterized differences between Nish and Ni-CE P3. Combining small-angle X-ray scattering (SAXS), cross-linking mass spectrometry (XL-MS), and NMR spectroscopy data suggests that the functional Nish P3 fluctuates between closed and open structures, whereas the defective Ni-CE P3 is essentially in the open state. The closed form brings the sequentially distant globular P_CTD and the IDRs of the N-terminal region (P_NTR) into proximity, suggesting that the formation of multiple weak interactions between them underpins diverse properties of P3. We also present in vitro data indicating that Nish P3 and Ni-CE P3 differ in their capacity to undergo LLPS, whereby Ni-CE P3 is highly defective, indicating that the closed form is required; this correlates with the association of the respective P3 proteins with MT bundles and nucleoli in cells, and with the capacity of virus to induce MT bundling in infected cells. Together, these data support multifunctionality of the protein deriving from specific spatial organization and interactions of IDRs and structured domains that can augment the basic modular organization of a protein to generate multiple diverse and functionally distinct “species” from a single sequence.

Results and Discussion

Host Cell Protein Interactions Differ between Nish and Ni-CE P3.

P1 displays a diffuse cytoplasmic phenotype, but the shorter P3 forms multiple host cell interfaces, including at MT bundles, nuclei, nucleoli, and nuclear bodies (5–9). We showed that the P gene is a pathogenicity-determining factor through the analysis of Nish RABV, which is pathogenic in mice, and a Nish-derivative strain, Ni-CE, that is more sensitive than Nish to interferon (IFN) and nonpathogenic. A recombinant strain, CE-NiP, in which the Ni-CE P gene is substituted for that of Nish displayed increased IFN antagonist function and pathogenesis, indicating critical roles of P protein (14, 15). Comparison of Nish and Ni-CE P3 indicated defects in the latter for MT association, nuclear localization, and antagonism of antiviral IFN signaling via signal transducers and activators of transcription 1 (STAT1) (6, 7). Ni-CE P3 differs from Nish P3 by five substitutions, L56P, L58P, L66P, F81P, and N226H (Fig. 1A). Importantly, replication functions are maintained (14), indicating that mutations are not globally detrimental but cause specific defects in host interactions and IFN/STAT1 antagonism, critical in pathogenesis. The effects of mutations on key interactions of P3 with host proteins have not been examined in vitro.

We recombinantly expressed and purified full-length Nish and Ni-CE P3, for the analysis of binding to tyrosine phosphorylated (pY) STAT1, the IFN-activated form of STAT1 which P protein binds as a critical element in IFN antagonism and disease (14, 16, 17), and LC8 that is implicated in MT-dependent nuclear trafficking of P protein (10, 18). Substitutions in Ni-CE P3 are separate from the STAT1 and LC8 binding sites, and STAT1-P protein binding is retained in immunoprecipitation assays from Ni-CE-infected cells at a level similar to that from CE-NiP-infected cells, but interactions have not quantitatively been examined (6, 16, 19, 20). Using isothermal titration calorimetry (ITC) (SI Appendix, Fig. S1 and Table S1), we showed that Nish P3 binds N-terminally truncated pY-STAT1 (residues 136 to 750 and N-terminally tagged with GB1) and LC8 with affinities (K_D) of 0.8 µM and 0.08 µM, respectively, consistent with previous investigations (16, 19). Ni-CE P3 showed a 25-fold reduction in affinity to pY-STAT1 (K_D, 19.3 µM), but a 5,000-fold loss of affinity to LC8 (K_D, 395.3 µM). These data are consistent with the reduced ability of Ni-CE to evade the innate immune system through reduced pY-STAT1 antagonism (6, 7), and with defective nuclear localization of Ni-CE P3, as LC8 binding facilitates nuclear import (10, 18). Loss of LC8 interaction could also contribute in part to defective MT interaction/bundling by Ni-CE P3. However, mutations disabling LC8 binding in the P3 protein of CVS strain RABV do not disable MT binding/bundling, so LC8 binding is not critical; furthermore, LC8 interaction appears to facilitate NLS-mediated nuclear import, while P3-MT binding/bundling is inhibitory, suggestive of a complex relationship of P proteins and MTs (10). The Ni-CE mutation, N226H, in the P_CTD is not in the proposed pY-STAT1 binding site but may have a modest off-target effect on pY-STAT1 binding, as previously observed for other mutations of the P_CTD (16). All Ni-CE mutations of P protein are outside of the proposed LC8 binding site (19) (Ser140 to Thr149 in IDR2; Fig. 1A), suggesting that conformational differences between Nish P3 and Ni-CE P3 within this region prevent the association with LC8 for the latter protein, and may also impact on STAT1 binding. Notably, previous analysis of nuclear localization of Nish and Ni-CE P3 indicated that Ni-CE mutations in the P_NTR and P_CTD impair nuclear localization consistent with effects on nuclear localization (NLS) and export (NES) sequences, despite these mutations not directly changing key residues identified as part of such sequences in the P3 isoform (7, 21). These data are consistent with a three-dimensional conformational organization of the domains of P3 being important to its multifunctionality. These observations are intriguing in light of previous observations (10, 11) that truncation of P1 protein to generate P3 not only removes functional sequences of the L-binding site and the P1 N-NES, but also causes de novo activation of the noncanonical N-NLS (encompassing residues 53 to 174), and a sequentially distant MT-association sequence, which has been mapped to the P_CTD.

RABV P3 Forms a “Closed” Structure through Interactions of IDRs and the P_CTD.

The availability of the purified Nish and Ni-CE P3 proteins provided the opportunity to examine the implied conformational organization of P3, and through comparison of Nish P3 with Ni-CE P3, relate conformational properties to pathogenic functions. P3 comprises two ordered domains, a helical dimerization domain (DD residues 92 to 131 through which the P protomer is a dimer), and a globular P_CTD (residues 183 to 297) (Fig. 1A). The DD is flanked on either side by two predicted [using IUPred2A (22)] IDRs (IDR1 and IDR2). SAXS analysis generated a profile of Nish P3 consistent with the expected dimeric protomer, with a mass of ~55 kDa showing a radius of gyration (R_g) of 39.2 Å and a particle diameter (D_max) of 145 Å (Fig. 1 B and C and SI Appendix, Fig. S2 and Table S2). A Kratky plot (Fig. 1D) of Nish P3 shows a rise to a plateau that is qualitatively characteristic of a partially disordered protein which is in complete agreement with the presence of ordered (P_CTD and DD) and disordered (IDR1 and IDR2) regions in Nish P3. Using the Coral program (23) within the ATSAS 3.2.1 software package (24, 25), we built a model of P3. Rigid-body modeling with a p2 symmetry was performed, using the crystal structures of the DD (PDB: 3L32) and the Nish P_CTD (PDB: 7T5H) and IDRs provided as linkers between the two structures. The model obtained represents the approximate conformation of the IDRs around the ordered domains and suggests a closed state where the IDR1 and IDR2 are in proximity to one another and to the P_CTD. The shape of Nish P3 was also determined Ab initio using DAMMIN (26) and DENSS (27) in agreement with this model (SI Appendix, Fig. S3), supporting the formation of a closed conformational state for Nish P3. To further model Nish P3 in solution, we used the ensemble optimization method (EOM) (28) to generate a pool of different possible conformations (Fig. 1E). The range of conformations for Nish P3 shows two distinct pools with a preference for the closed or compact state.

To directly investigate the structural topology and predicted proximity of the domains, we used chemical XL-MS with disuccinimidyl sulfoxide (DSSO) cross-linker (29), which covalently links the sidechain amine groups of lysine residues that are within 25 Å of each other (30, 31). 37 intraprotomer cross-links were detected (SI Appendix, Table S3). There are two lysine residues in IDR1 at positions 63 and 72, of which Lys63 formed cross-links to five lysine residues of the P_CTD and five of IDR2 (Fig. 2A and SI Appendix, Table S3). Of the five lysines of the IDR2, three cross-linked to five lysines in the P_CTD. These XL-MS data suggest a diversity of conformational states and indicate that IDR1 is proximal to both IDR2 and the P_CTD, consistent with our SAXS model (Fig. 1E) and SI Appendix, Figure S3. Two cross-links (Lys282-Lys214 and Lys282-Lys256) detected within the P_CTD were in consensus with the distance restraint for intradomain cross-linking (< 27 Å) when measured as nonlinear distances (32) over the surface of the solved structure of the Nish P_CTD (21) (Fig. 2D), whereas the remainder were much greater, indicating that some cross-links may be between the two subunits of the same P3 dimer. However, not all lysine residues in the P_CTD (Fig. 2D) participated in cross-linking with the disordered regions, suggesting a limitation on the formation of specific weakly interacting interfaces between the P_CTD and IDRs.

Fig. 2.

The P_CTD of Nish P3, but not Ni-CE P3, interacts with both IDR1 and IDR2. (A and B) XL-MS patterns for Nish P3 (A) and Ni-CE P3 (B) plotted on strip diagrams of the domain structure of P3 [IDR1 and IDR2 are intrinsically disordered regions; DD, dimerization domain; CTD, C-terminal domain (P_CTD)]. The cross-links between IDR1 and the CTD observed for Nish P3 are absent in Ni-CE P3, while the remaining cross-links are similar (SI Appendix, Table S3). (C) NMR-monitored titration of 100 μM ¹⁵N-labeled P_CTD with 200 μM unlabeled CTD-truncated P3 (residues 53 to 185) (black) and a titration of 100 μM ¹⁵N-labeled CTD-truncated P3 (residue 53 to 185) with 200 μM unlabeled P_CTD (blue). The titrations show average chemical shift deviations (Δδ) for residues that are consistent with the XL-MS data in (A). (D) Structure of the P_CTD (PDB: 7T5H) highlighting the side chains of the lysine residues. Those colored cyan and annotated are Lys shown to participate in cross-links whereas those in orange do not cross-link suggesting that the P_CTD shows a preferential orientation to the IDRs. Residues sensitive to the NMR titration are colored in dark blue and show proximity to Lys that cross-links to IDR1 or IDR2.

To confirm that the P_CTD interacts with the IDRs of P_NTR, we conducted ¹⁵N-HSQC-monitored NMR titration experiments of ¹⁵N-labeled P_NTR with unlabeled P_CTD and vice versa. Amide resonances of residues of the ¹⁵N-labeled P_NTR sensitive to the addition of the P_CTD (Fig. 2C and SI Appendix, Fig. S4) (where δ_av > 0.01 ppm) comprise 138 to 169 of IDR2, which agrees well with the XL-MS data of the P_CTD with IDR2 although few interactions were observed to assigned residues in IDR1 55 to 88 (Fig. 2A), suggesting these latter interactions may only occur in an intramolecular context. In the reverse titration, residues of ¹⁵N-labeled P_CTD that showed δ_av >0.01 ppm were from the region comprising the C-terminal helix-7 (Asp290, Asn292, and Arg293), helix-5 (Cys261 and Val262) that packs against helix-7, and residue Leu257 N-terminal to helix-5 (Fig. 2D). These residues are surrounded by Lys214, 256, and 282 that cross-link to IDR1 and IDR2 in the XL-MS data. Other residues that showed δ_av >0.01 ppm were Gly221 (near Lys214 and 282 that show XL-MS-observed cross-links), Val238 (near Lys242), and Lys273. Together, these data support the formation of a “closed” conformation which is enabled by the binding interface between the sequentially distant IDR1, IDR2, and P_CTD.

Structures of P3 from Pathogenic and Nonpathogenic RABV Differ.

Based on the above data and previous results from cell-based assays indicating that MT interaction and nuclear import of P3 might be regulated by interactions of the N-terminal and C-terminal regions (10, 11), we predicted that the structure of Ni-CE P3 would be altered. Indeed, SAXS analysis of Ni-CE P3 (Fig. 1 B and C) indicated that this protein is significantly elongated compared to Nish P3 supported by the higher R_g of 43.07 Å and D_max of 168 Å (Fig. 1 and SI Appendix, Fig. S2 and Table S2) and rigid-body modeling which indicates that IDR1 is not in proximity to the P_CTD or IDR2. The Kratky plot (Fig. 1D) suggests that Ni-CE P3 is still partially disordered, but may not be as conformationally diverse as Nish P3. Ab initio modeling of Ni-CE P3 suggests that it is also less compact than Nish P3 (SI Appendix, Fig. S3). The pool of conformations produced by EOM analysis (Fig. 1F) continues to show two distinct pools similar to Nish P3, but the preference for a compact state is significantly reduced, with a greater proportion of “open” and elongated states with higher R_g and D_max. Consistent with the SAXS data, XL-MS showed specific loss of lysine–lysine cross-links between IDR1 and the P_CTD, and between IDR1 and IDR2 (Fig. 2B), but not between IDR2 and the P_CTD. Thus, the mutations of Ni-CE P3 appear to abolish the interactions from IDR1 to both IDR2 and the P_CTD.

Of the mutations in Ni-CE-P3, N226H appears sufficient to significantly inhibit nuclear import and MT association (6, 7). Therefore, we assessed the conformational effects of N226H mutation on P3 using SAXS. SAXS analysis suggested that N226H P3 is also significantly elongated compared to Nish P3 with an R_g of 42.89 Å and D_max of 152.82 Å (Fig. 1 and SI Appendix, Fig. S2 and Table S2), much higher than the Nish P3 but slightly lower than Ni-CE P3. Ab initio modeling (SI Appendix, Fig. S3) and EOM-derived pool of conformations for N226H P3 (Fig. 1G) also suggest a more conformationally diverse ensemble with no preference for a compact state, very similar to the models and EOM profile of Ni-CE P3. Sample preparation of N226H P3 for XL-MS experiments unfortunately resulted in precipitation, presumably due to the DMSO used to dissolve DSSO, suggesting that in the absence of the IDR1 mutations, P3-containing N226H alone is less stable than Nish and Ni-CE P3.

Thus, Ni-CE P3 and Nish P3 containing only the N226H mutation predominantly form an open structure that is defective for specialized phenotypes of P3 correlating with loss of pathogenic function, suggesting that the closed structure of Nish P3 is the active multifunctional form. As Nish P3 can exist in several conformations, it is likely that switching to the open form may provide a dynamic regulatory mechanism to turn functions on or off. These observations provide insight into how the N-terminal end of P protein, through truncation of IDR1 to form P3, can regulate functions, such as MT association and bundling, which appears to be dependent on the P_CTD, and nuclear trafficking, which involves activation of the extended noncanonical N-NLS. Our data suggest that association of IDR1 and IDR2 of the P_NTR is essential for the activity of the N-NLS. Notably, our findings that LC8 binding, which facilitates nuclear import, also appears to be regulated through conformational changes are indicative of potential coordinated effects to regulate P3 nuclear trafficking.

Thus, rather than comprising a simple mosaic of independent interaction sequences, key functions of P3 appear to derive from the three-dimensional structural organization of the protein that juxtapositions regulatory regions, the function of which can be modified by mechanisms such as truncation, mutations between RABV strains, and phosphorylation (6, 7, 10–12, 21). This concept refines the conventional multimodular model for P3 (and other such multifunctional proteins) where addition or deletion of discrete modules produces a composite phenotype, indicating that broader conformational organization of domains and disordered regions enables a higher-order regulation of the phenotype, enabling the generation of diverse protein species. This may endow viruses a flexibility to generate multiple functionally unique proteins from limited genomic material. Notably, while our data highlight the importance of the closed structure, and indicate a “dysfunctional” nature of an open structure, it seems likely that the flexibility of the IDRs permits a dynamic conformation, potentially switching between these states and regulated by mechanisms such as posttranslational modifications or intermolecular interactions.

Nish P3 but Not Ni-CE P3 Can Form Liquid Bodies In Vitro.

RABV, and a number of other viruses (33–37), form replication centers through LLPS (36, 38), implicating phase separation in viral infection/replication (39). For RABV, these replication centers (known as Negri bodies) depend on both the viral N and P proteins, where the amino terminal portion of IDR2 of P protein was found essential for LLPS (36, 38). Several studies indicate that P3 interacts with cellular structures formed by LLPS including nucleoli, and nuclear bodies, where P3 colocalizes with molecular partners of P protein such as nucleolin and PML protein. P3 expression also impacts on LLPS-dependent structures, increasing MT bundling (indicating that P3 induces bundles, or targets and stabilizes MT bundles) and increasing PML body size, similar to effects of infection (5, 10, 40), and impairs rRNA biogenesis in nucleoli (5–7, 41). As IDRs are often implicated in the interaction and formation of MLOs (25–28), we considered that P3 may undergo LLPS facilitating cellular MLO association. Since conformational differences affecting the IDR1/2 and P_CTD interactions correlate with a defect in MT bundling by Ni-CE P3 (6), we hypothesized that LLPS propensity would differ between Nish P3 and Ni-CE P3.

We tested the potential of Nish P3 and Ni-CE P3 to form liquid droplets in vitro using light microscopy (42). Analysis of images (three fields of view for each protein) indicated that at 12.5 to 100 μM Nish P3 produced 10 to 210 liquid droplets, respectively, whereas Ni-CE P3 was severely impaired (SI Appendix, Fig. S5). Thus, impaired MT association/bundling and IFN antagonist function, which is dependent on MTs (6, 10), may correlate with defective droplet formation of Ni-CE P3. A Nish P3 chimera containing only the NTR mutations of Ni-CE P3 (L56P, L58P, L66P, and F81P; NTRm) showed LLPS similar to Nish P3, consistent with the retention of MT association/bundling observed in cells (6). However, P3 containing only N226H (at 100 μM) showed reduced LLPS, similar to Ni-CE P3, consistent with a significant defect in MT association and IFN antagonist function in cells, suggesting that this latter mutation antagonizes the capacity of Ni-CE P3 to undergo LLPS, correlating with defective functions.

Together, these data indicate that substitutions in Ni-CE P3 effect broad changes in the conformation of Ni-CE P3, that impact on the formation of functional interfaces, and so concurrently affect several cellular phenotypes. Thus, P3 function appears to result from the presence and organization of sequences within domains, the spatial organization of these domains, and, potentially, the general capacity to undergo LLPS.

Ni-CE P3 Is Defective for Targeting Nucleoli.

P3 targets several cellular MLOs in addition to MT bundles, such that defects in Ni-CE P3 structure may affect the formation of the broader host interface of P3. P3 of the CVS strain of RABV can enter nucleoli and colocalize and physically interact with nucleolin (40). Using CLSM analysis of live cells expressing Nish P3, we confirmed accumulation into nuclei (Fig. 3A), but Ni-CE P3 remained largely cytoplasmic, as described previously (7), which appears to be due to the effects of NTR mutations in inhibiting the function of the N-NLS and N226H in inhibiting C-NLS and enhancing C-NES function; based on our data (above), reduced interaction with LC8 by Ni-CE P3 may also contribute (14, 43). Nish P3 also clearly accumulated into nucleoli, but no accumulation was evident for Ni-CE P3 (Fig. 3B), due to defective nucleolar localization and/or nuclear import. To enhance nuclear localization of Ni-CE P3, we treated cells with leptomycin B (LMB, a specific inhibitor of exportin 1, which consequently inhibits P3 nuclear export) (12, 44). LMB increased nuclear localization of Ni-CE P3, although it did not accumulate to similar levels as Nish P3, due to defects in the NLSs (7); furthermore, little to no nucleolar accumulation was evident (Fig. 3A). Quantitative image analysis to calculate the ratio of nucleolar to nuclear fluorescence (Fnu/n) in n ≥ 21 cells for each condition indicated an Fnu/n greater than 1 for Nish P3, but of c. 1 for Ni-CE P3, consistent with loss of nucleolar accumulation (Fig. 3B). Nish P3 containing only the substitutions within the Ni-CE P_NTR [L55P, L56P, L66P, and F81P; NTRm, (Fig. 1A)] was impaired for nuclear accumulation both in LMB-treated and -untreated cells as expected due to effects on the N-NLS (7), but accumulation in the nucleolus was observed, and quantitation indicated a similar level of accumulation compared with nucleoplasmic protein as observed for Nish P3 (Fig. 3B). N226H mutation alone resulted in a significant reduction in nuclear accumulation with a substantial increase by LMB (Fig. 3A), consistent with a major effect of N226H being through the indirect enhancement of C-NES-driven nuclear export (7). Nevertheless, little to no nucleolar localization was evident with or without LMB treatment (Fig. 3B). This is consistent with the differing potential of N226H-containing P3 proteins for phase separation in vitro (SI Appendix, Fig. S5), and with previous data using CVS P protein that showed truncates, comprising the P_CTD but not the P_NTR, can localize within nucleoli (40).

Fig. 3.

Nish P3, but not Ni-CE P3 and N226H P3, associates efficiently with MLOs in vivo. (A) HeLa cells transfected to express GFP-fused Nish, Ni-CE, NTRm, and N226H P3 proteins were imaged live using CLSM. Cells were either untreated or treated with LMB or Taxol. (B) Images such as those in B were used to determine the nucleolar-to-nuclear fluorescence ratio [mean Fnu/n ± SEM, n ≥ 21 cells for each condition; dotted line indicates Fnu/n of 1 (no accumulation)]. (C) Images such as those in B were used to determine the percentage of cells with clear P3-MT association, apparent as filamentous organization (n ≥ 32 cells for each condition). ns, nonsignificant; ****P < 0.0001.

Analysis of P3-MT association, evident as cytoplasmic filamentous P3 that is enhanced by the MT-stabilizing agent Taxol (Fig. 3 A and C), confirmed strong inhibition by N226H mutation, with only minor effects of P_NTR mutations. Notably, the effects paralleled those on nucleolar accumulation (Fig. 3 A and B), consistent with substitutions that result in an open structure for Ni-CE P3 (Fig. 1) and impair potential to undergo LLPS (SI Appendix, Fig. S5).

RABV Infection Induces MT Bundling Which Is Defective for Virus Expressing Ni-CE P3.

Collectively, our data suggest that the conformational organization of P3 is important to MT interaction/bundling, enabling functions in immune evasion and pathogenesis (5, 6, 40). Although an association of P3 with MLOs has been reported in several studies, confirmation of interactions such as with MT bundles in infected cells is lacking, largely due to the lack of reagents that can distinguish the highly expressed longer P1 and P2 isoforms, that do not bind MTs, from the less abundant, P3 (6). The availability of viruses that are identical except for the P gene (Fig. 4A): Ni-CE; CE(NiP) (Ni-CE virus in which the P gene is substituted for the P gene of Nish) and CE(NiP-N₂₂₆H) (Ni-CE virus in which the P gene of Nish contains the N226H mutation), together with our finding that P3-MT association can be indirectly measured by quantifying MT bundling by superresolution microscopy, provided the opportunity to examine the effects of P gene products on MTs during infection (6, 21). We reasoned that P isoform–MT interaction during infection might be detectable as bundling, which would be predicted to differ between viruses expressing Nish or Ni-CE P genes (6).

Fig. 4.

MT bundling is evident in cells infected by RABV carrying the Nish P gene but not RABV carrying the Ni-CE P gene. (A) Schematic representation of the genomes of viruses used; genes from Nish and Ni-CE are in black and white, respectively. The P gene of Ni-CE is substituted for the P gene of Ni in CE(NiP), and for a mutated version of the Ni-P gene containing N226H in CE(NiP-N₂₂₆H). (B) dSTORM images of immunostained β-tubulin in SK-N-SH cells infected with the indicated virus; MTfds for the indicated filaments are shown below the corresponding images. (C and D) Tukey box plots (C) and frequency distribution (D) of MTfds calculated for each virus (n = 504 [CE(NiP)], 516 [CE(NiP-N₂₂₆H)], 284 [Ni-CE], and 322 [Mock]; measurements are from ≥7 cells for each virus across two independent assays).

We infected SK-N-SH neuroblastoma cells with Ni-CE, CE(NiP), or CE(NiP-N₂₂₆H) viruses. These viral strains were used previously to demonstrate that the MT interaction/IFN antagonist function of Ni-CE or N226H P3 proteins expressed alone correlates with viral pathogenesis (6). We developed a workflow to fix and inactivate infected cells while maintaining MT structures for dSTORM (direct stochastic optical reconstruction microscopy), a technique that can sensitively quantify P3–MT interaction via bundling (6, 21) (Fig. 4B). Substantial bundling was evident in CE(NiP)-infected cells, but this was reduced in CE(NiP-N₂₂₆H)-infected cells, and, to an even greater extent, in Ni-CE-infected cells (6, 21). Analysis of median diameters of MT features (MTfd) confirmed that MT bundling in cells infected with CE(NiP) was significantly greater than that in cells infected with CE(NiP-N₂₂₆H) or Ni-CE (Fig. 4 C and D). Thus, MT bundling occurs during infection and is impaired in cells infected by virus containing substitutions from the Ni-CE P gene, including N226H; these data correlate with observations for cells expressing the P3 proteins alone (6), indicating that P3–MT interactions occur and affect MT structures during infection.

Conclusion

Rather than comprising a simple mosaic of independent interaction sequences, key functions of P3 appear to derive from the three-dimensional structural organization of the protein that juxtapositions regulatory regions, which can be regulated by mechanisms such as truncation and mutations between RABV strains. For instance, nuclear targeting usually involves short mono-partite or bi-partite motifs of basic residues, but the nuclear localization activity of the P_NTR appears to use a nonclassical nuclear localization sequence, requiring distant residues and so has been speculated to depend on structural organization. Our finding that disruption of the association of IDR1 and IDR2 in Ni-CE P3 and N226H P3 antagonizes nuclear accumulation is consistent with such an organization. Similarly, MT association appears to involve an interplay of the P_NTR and P_CTD, including dimerization by the NTR-localized DD, and regulation by truncation of the N terminus, and again is inhibited by disruption of the interaction of the IDR1 with downstream elements of the protein. This concept refines the conventional multimodular model for P3 (and other such multifunctional proteins) where addition or deletion of discrete and independent modules produces a composite phenotype, indicating that broader conformational organization of domains and disordered regions enables a higher-order regulation of the phenotype, enabling the generation of diverse protein species. This may endow viruses a flexibility to generate multiple functionally unique proteins from limited genomic material.

Materials and Methods

Recombinant Protein Expression and Purification.

Expression and purification of P3 proteins (wild type and mutants) P_NTD and P_CTD, LC8, and phosphorylated N-terminally truncated STAT1 are described in SI Appendix.

Mammalian Cell Culture.

Plasmids used for mammalian expression and cell-culturing protocols are described in SI Appendix.

ITC.

Thermodynamic analysis of the interaction between P3 with pY-STAT1 and LC8 was conducted with a MicroCal iTC200 (GE Healthcare). The cell was loaded with pY-STAT1 or LC8 and the syringe with P3. Experimental data were fitted to single binding site models using MicroCal ORIGIN software. Further details are in SI Appendix.

SAXS.

SAXS measurements were conducted at the Australian Synchrotron SAXS/WAXS Beamline equipped with a coflow system to avoid radiation damage and enable higher X-ray flux (11,500 eV) and with an in-line SEC to limit protein sample dilution (45–47). Collected SAXS data were reduced using the Scatterbrain software and analyzed by CHROMIXS (48) and the ATSAS 3.2.1 software package (24, 25). Complete details of the SAXS experiments and their analysis are described in SI Appendix.

Cross-Linking-Mass Spectrometry.

For cross-linking-mass spectrometry (XL-MS), purified P3 protein samples were mixed with 100-fold excess of DSSO cross-linker (29) dissolved in dimenthyl sulfoxide (DMSO). After the cross-link reaction was terminated, the cross-linked proteins were treated with trypsin. LC MS/MS was carried out using the Fusion Lumos Orbitrap mass spectrometer with the FAIMS Pro source (Thermo Fisher, USA). To identify the cross-linked peptides, the MS2CID-MS3HCD (MS2-MS3) workflow was used. The cross-linked peptides were identified using the XlinkX (30) node-implemented Proteome Discoverer 2.3 (Thermo Fisher Scientific). Full details of the sample preparation, workflow, and analysis are described in SI Appendix.

NMR Spectroscopy.

2D ¹⁵N,¹H HSQC-monitored titrations were acquired on a Bruker Avance 700 MHz NMR spectrometer equipped with a TCI cryoprobe. 100 µM ¹⁵N-labeled P_CTD was titrated with 100 and 200 µM unlabeled P_NTR. Similarly, 100 µM ¹⁵N-labeled P_NTR was titrated to 200 µM unlabeled P_CTD. Details of experiments for assignment of the IDR1 and IDR2 of the P_NTD of P3 are described in SI Appendix.

In Vitro LLPS Assay.

In vitro phase separation assays were based on published protocols (49). Further details are in SI Appendix.

Confocal Laser Scanning Microscopy (CLSM) and Image Analysis.

CLSM of cells was performed as previously described (6, 7), with analysis using ImageJ freeware software as previously described to quantify nuclear and nucleolar fluorescence or % of cells with detectable P3-MT association (see SI Appendix for details).

Virus Infections and dSTORM Analysis.

SK-N-SH neuroblastoma cells were infected with RABV and variants for 40 h (MOI = 3) prior to fixation and processing for immunofluorescent staining for β-tubulin as described in SI Appendix. dSTORM imaging of virus-infected cells was performed using a home-built widefield superresolution microscope setup as previously described (50). Further details are in SI Appendix.

Statistical Analysis.

For Fnu/n analysis, GraphPad Prism version 9.2.0 (GraphPad Software, San Diego, California USA) was used to calculate P values using Student’s t test (unpaired, two tailed). For dSTORM analysis, P values were determined using one-way ANOVA (Kruskal–Wallis test with Dunn’s multiple comparisons test). Significance is represented using ***P ≤ 0.0001, ns = not significant.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix. Assignments for the IDR1 and IDR2 of Nish P3 are deposited at the Biological Magnetic Resonance Data Bank under the accession number 51617 (51). SAXS data are deposited in the Small-Angle Scattering Biological Data Bank with accession codes SASDQG4 (Nish P3) (52), SASDQH4 (Ni-CE P3) (53), and SASDQJ4 (N226H P3) (54).