Structural Description of the Nipah Virus Phosphoprotein and Its Interaction with STAT1

Abstract

The structural characterization of modular proteins containing long intrinsically disordered regions intercalated with folded domains is complicated by their conformational diversity and flexibility and requires the integration of multiple experimental approaches. Nipah virus (NiV) phosphoprotein, an essential component of the viral RNA transcription/replication machine and a component of the viral arsenal that hijacks cellular components and counteracts host immune responses, is a prototypical model for such modular proteins. Curiously, the phosphoprotein of NiV is significantly longer than the corresponding protein of other paramyxoviruses. Here, we combine multiple biophysical methods, including x-ray crystallography, NMR spectroscopy, and small angle x-ray scattering, to characterize the structure of this protein and provide an atomistic representation of the full-length protein in the form of a conformational ensemble. We show that full-length NiV phosphoprotein is tetrameric, and we solve the crystal structure of its tetramerization domain. Using NMR spectroscopy and small angle x-ray scattering, we show that the long N-terminal intrinsically disordered region and the linker connecting the tetramerization domain to the C-terminal X domain exchange between multiple conformations while containing short regions of residual secondary structure. Some of these transient helices are known to interact with partners, whereas others represent putative binding sites for yet unidentified proteins. Finally, using NMR spectroscopy and isothermal titration calorimetry, we map a region of the phosphoprotein, comprising residues between 110 and 140 and common to the V and W proteins, that binds with weak affinity to STAT1 and confirm the involvement of key amino acids of the viral protein in this interaction. This provides new, to our knowledge, insights into how the phosphoprotein and the nonstructural V and W proteins of NiV perform their multiple functions.

Significance

The intrinsically disordered phosphoprotein of Nipah virus orchestrates viral genome transcription, replication, and encapsidation and shares a long N-terminal disordered region with the nonstructural V and W proteins involved in blocking the interferon responses. By combining multiple biophysical methods, including x-ray crystallography, NMR spectroscopy, and small angle x-ray scattering, we characterized the structure and flexibility of this protein, identified several putative short linear motifs of interaction with viral and cellular partners, and investigated the interaction with one of them. The protein appears as a nanometer-sized octopus comprising a central tetrameric core and displaying binding sites on its flexible arms to recruit multiple viral and cellular partners, thus providing new, to our knowledge, insights into how this protein fulfills its multiple functions.

Introduction

Nipah virus (NiV) is one of the most pathogenic and deadly viruses infecting humans (1). It is an emergent zoonotic virus that naturally infects fruit bats in South-East Asia but that can also infect other hosts, including pigs, dogs, cats, horses, and humans, in whom it becomes highly pathogenic. Since its discovery in 1998 in Malaysia, NiV has caused regular outbreaks of severe human encephalitis and respiratory illness with a fatality rate ranging from 40 to 90%, mainly in Bangladesh and, more recently, in the Philippines and India (1,2). With its ability to transmit from human to human and the lack of vaccine or efficient treatment, NiV has the potential to cause a pandemic and to be used in biological warfare (3). It is consequently classified as a biosafety level 4 pathogen.

NiV is an enveloped virus with a nonsegmented negative-sense RNA genome that belongs to the order Mononegavirales and the family Paramyxoviridae, although several features in its genome organization and in its protein sequences led to its classification into a new genus, Henipavirus, with the highly pathogenic Hendra virus and the nonpathogenic Cedar virus (4). The multiplication of NiV in a suitable host cell, like that of all members of the Mononegavirales, requires both 1) transcription of the single-stranded RNA genome into monocistronic, capped, and polyadenylated messenger RNAs (positive-sense RNA) that can be used by the cellular ribosomes to generate new viral proteins and 2) replication of the negative-sense RNA genome into positive-sense antigenome copies, which are then used to generate a progeny of negative-sense genomes. The enzymatic machinery that carries out these activities is conserved among all members of the Mononegavirales and is unique in biology, and it thus represents an attractive target for the development of much-needed antiviral compounds. It consists of a megadalton-sized multimolecular complex located in the core of the viral particle and timely released in the cytoplasm of the host cell (5,6). This transcription/replication complex is composed of the genomic RNA and of three viral proteins: the nucleoprotein (N), the phosphoprotein (P), and the large RNA-dependent RNA polymerase (L). With the help of the phosphoprotein acting as a chaperone, the nucleoprotein assembles into linear homopolymers that coats both negative-sense genomic and positive-sense antigenomic RNAs, forming long, helical nucleocapsids (7). These nucleocapsids constitute the biologically active templates used by the polymerase complex, which consists of the L polymerase and the phosphoprotein acting as a noncatalytic cofactor. Structural and dynamical studies of the unassembled nucleoprotein (N⁰) in complex with the phosphoprotein (8, 9, 10, 11) and of nucleocapsids (7,12) have started to unravel the mechanisms of assembly and remodeling of the nucleocapsids during the transcription and replication processes (13). The multidomain L polymerase catalyzes RNA synthesis but also the capping, methylation, and polyadenylation of the viral messenger RNAs (14). The underlying mechanisms of the functioning of this polymerase remain largely to be deciphered (15), but its structure in complex with its cofactor, the phosphoprotein, is being revealed by recent cryo-electron microscopic reconstructions of the complex from various Mononegavirales (16, 17, 18, 19), including a first low-resolution reconstruction of the NiV polymerase (14). The third essential component, the phosphoprotein, is thus a multitasking protein that orchestrates the interplay between the different components of the transcription/replication complex. The phosphoprotein takes its name from early observations showing that it is the most heavily phosphorylated viral protein in cells infected by different members of the Mononegavirales (20). In NiV phosphoprotein, residues Ser224 and Thr239 are the major phosphorylation sites (21), and a binding site for Polo-like kinase 1 is located near residue 130 (22), but as for other Mononegavirales, the role of these phosphorylations in the viral replication cycle remains elusive (23).

All phosphoproteins of the Mononegavirales share the properties of binding the L polymerase, the unassembled nucleoprotein, and the nucleocapsid through independent structural modules (13). In this way, the phosphoprotein promotes the initiation of RNA synthesis, confers processivity to the polymerase, maintains the unassembled nucleoprotein in an RNA-free and soluble form, and delivers the latter to the site of RNA synthesis to encapsulate the nascent RNAs (9,10,12,24, 25, 26, 27, 28). Despite common functions, similarities in the phosphoprotein sequence are rarely found beyond the level of the family and in some cases of the genus (29,30). However, all phosphoproteins of the Mononegavirales are multimeric and share a common architecture with a long N-terminal intrinsically disordered region (P_NTR) and a C-terminal region (P_CTR) that includes the multimerization domain (P_MD), a linker region (P_CT-LINK), and most often a folded domain at the C-terminal end (P_XD) (Fig. 1 B; (31,32)). They also share a rather common localization of functional modules (14,32, 33, 34), and thus, NiV phosphoprotein exhibits a structural architecture and an organization of the functional modules involved in binding the nucleoprotein and the polymerase that are typical of the family Paramyxoviridae, although the protein of henipaviruses is appreciably longer than that of the other genera of Paramyxoviridae (Fig. 1 B; (25,33,35)): 1) the N-terminal disordered region of more than 400 amino acids (aa) contains a short region (aa 1–38) at the N-terminal end that acts as a molecular recognition element or MoRE (i.e., a region that is partially disordered in isolation but that adopts a stable conformation upon binding to a partner (36)) for recruiting and chaperoning the RNA-free unassembled nucleoprotein (N⁰) required for the assembly of new nucleocapsids during genome replication (9). 2) The tetramerization domain of NiV (aa 475–578) (37) is involved in the interaction with and in the function of the L polymerase (27,34,38). 3) A folded domain at the C-terminal end of the protein, named the X domain (P_XD) (aa 655–709), binds to an MoRE in the C-terminal disordered tail of the nucleoprotein (N_TAIL) (39). 4) A third region of weak interaction with the nucleoprotein that has recently been identified in measles virus phosphoprotein appears to be conserved in NiV phosphoprotein (aa 330–350) (10). Despite all this knowledge, we are still lacking a structural characterization of the full-length NiV phosphoprotein.

Figure 1.

Modular architecture of Nipah virus phosphoprotein. (A) Consensus disordered prediction is shown. The D-score was calculated as described in (32). The threshold to distinguish between the ordered and disordered regions is set at 0.5. The shaded areas indicate the known positions of the folded tetramerization domain and X domain. (B) The modular organization of the NiV phosphoprotein is shown. The upper part shows the structural organization of the phosphoprotein encoded in the main open reading frame (0 ORF). Boxes indicate the localization of folded domains, undulated lines the localization of predicted MoREs, and lines the localization of intrinsically disordered regions (IDRs 1–3). The green dots show two identified phosphorylation sites. In the middle part, the lines show the location of the known functional regions. In the lower part, the blue lines show the position of the C gene and of the regions coding for the specific parts of the V and W proteins in the alternative ORFs (+1 ORF and +2 ORF). (C) The divide-and-conquer strategy is shown. The lines indicate the position of the different fragments of the phosphoprotein used in this study. To see this figure in color, go online.

In addition, besides its roles in the transcription and replication of the viral genome, the phosphoprotein of many Mononegavirales also plays roles in hijacking cellular machineries and in counteracting the host immune responses (40). As in other Paramyxoviridae, the NiV P gene codes for multiple proteins through the use of alternative open reading frames (ORFs), and these additional proteins act as virulence factors, in particular by directly targeting the signal transducer and activator of transcription (STAT) proteins and blocking the interferon signaling pathway (40). The NiV P gene encodes three proteins, the phosphoprotein and the nonstructural V and W proteins. The gene contains an editing site in the part coding for the N-terminal region of the phosphoprotein, where the polymerase stutters during transcription and incorporates one, two, or more pseudotemplated Gs, resulting in shifts of the reading frame (41). The different messenger RNAs generated through this process lead to the expression of either the phosphoprotein or of one of the nonstructural V (+1 ORF) or W (+2 ORF) proteins, which share their 405 N-terminal residues with the phosphoprotein but have specific C-terminal moieties of 51 and 44 amino acids, respectively (Fig. 1 B; (41)). In addition, a second gene coding for the C protein is overprinted on the initial region of the P gene in the same reading frame as the V protein (+1 ORF) (Fig. 1 B) and is expressed from the three different mRNAs by a leaky ribosome scanning mechanism (42). However, in contrast to some other Paramyxoviridae, NiV phosphoprotein, V and W proteins block interferon signaling by preventing the phosphorylation of STAT1 and by sequestering STAT1 and STAT2 into nonfunctional complexes (43). The phosphoprotein and the V protein retain STAT proteins in the cytoplasm, whereas the W protein sequesters them in the nucleus. The N-terminal part comprising residues between 100 and 300, shared by the three proteins, is critical for interferon evasion. The binding site for STAT1 was first localized between residues 100 and 160 (44,45) and then further mapped down to residues 114–140 (6). Mutation of different residues in this region abrogate interferon inhibition (6,46). The interaction with STAT2 requires the entire 100–300 region, in particular a motif between residues 230 and 237, and the prior binding of STAT1 (44). Furthermore, the region between residues 174 and 192 contains a CRM1-dependent nuclear export signal, which allows the V protein to shuttle between the nucleus and cytoplasm and remove STAT1 out of the nucleus (43,47, 48, 49), whereas the W protein maintains phosphorylated STAT1 in the nucleus through the presence of a nuclear localization signal in its unique C-terminal moiety (6,45). Finally, the region around residue 130, overlapping with the STAT1 binding site, interacts with polo-like kinase 1, and this interaction results in the phosphorylation of the V protein (22).

Here, we investigate the structure of full-length NiV phosphoprotein (P_FL). We determine its oligomeric state by biophysical methods and demonstrate that the multimerization domain is the only region of the phosphoprotein involved in its tetramerization. We solve the crystal structure of the tetramerization domain and use NMR spectroscopy to investigate the structural properties of the long N-terminal region, of the C-terminal linker predicted to be intrinsically disordered, and of the C-terminal X domain. We demonstrate the independence of the different N-terminal chains in the tetrameric protein and identify several regions with transiently populated secondary structure elements, which could be MoREs for different viral or cellular partners. We generate an ensemble of conformers of full-length phosphoprotein that incorporates the crystal structure of the tetramerization domain and a homology model of the X domain and that simultaneously accounts for NMR and small angle x-ray scattering data from three different constructs of the protein. Finally, we investigate the interaction between STAT1 and a fragment of 100 amino acids derived from the N-terminal region of the phosphoprotein. NMR spectroscopy is used to validate the binding site in the phosphoprotein, whereas isothermal titration calorimetry and cell-based assays are combined with site-directed mutagenesis to map important residues for the interaction.

Materials and Methods

Disorder metaprediction

The location of intrinsically disordered regions (IDRs) in NiV phosphoprotein was predicted by submitting its amino-acid sequence to 16 different algorithms accessible through web servers and by calculating a consensus prediction as described previously (32,50). The calculated disorder score (D-score) varies between 0 and 1, and generally, regions with a D-score lower than 0.5 are intrinsically disordered, whereas regions with a score higher than 0.5 are either folded domains (50, 51, 52) or potential molecular recognition elements for partner proteins (53).

Protein expression and purification

On the basis of our predictions, different fragments of the phosphoprotein (Uniprot: Q9IK91) were cloned between the NcoI and BamHI restriction sites of a pET28 vector such that each construct included a C-terminal His₆-tag with a Leu-Glu linker. For the constructs lacking the N-terminal end of the protein, with the exception of the P_655–709 construct that starts at Met655, a Met residue or a Met-Ala dipeptide (construct P_173–240) was introduced at the N-terminal end of the fragment. Three single mutants of P_92–190 (Y116A, G120E, G125E) and a triple mutant (H117A/Q118A/H1119A) were generated by reverse PCR using the QuickChange mutagenesis protocol (Agilent, Santa Clara, CA). All constructs were verified by DNA sequencing.

The proteins were expressed in Escherichia coli BL21-RIL. Cells were grown in Luria Bertoni medium at 37°C until the optical density at 600 nm reached a value of 0.6; protein expression was then induced by adding 1 mM isopropyl β-D-1-thiogalactopyranoside, and cells were allowed to grow for 3 h. For producing the selenomethionine-substituted form of the P_471–580 construct, the bacterial cells were grown in minimal medium, and SeMet was added before induction of the heterologous protein expression. In each case, the cells were harvested by centrifugation, and the pellet was resuspended in 20 mM Tris-HCl buffer at pH 7.5 containing 150 mM NaCl (buffer A) and EDTA-free protease inhibitor cocktail (Roche, Basel, Switzerland). The cells were broken by sonication, and the soluble fraction was loaded onto a Ni²⁺ column equilibrated in buffer A. After washing the column with buffer A containing 20 mM imidazole, the protein was eluted with buffer A containing 300 mM imidazole. The protein was further purified by size exclusion chromatography on a Superdex S75 (GE Healthcare, Chicago, IL) equilibrated in buffer A or in 20 mM Bis-Tris buffer at pH 6.0 containing 150 mM NaCl, 50 mM arginine, 50 mM glutamate, and 0.5 mM TCEP (buffer B) for NMR experiments or in 20 mM sodium phosphate buffer at pH 7.5 containing 150 mM NaCl (buffer C) for isothermal titration calorimetry (ITC) experiments. Each protein preparation was checked by SDS-PAGE and SEC-MALLS (see below).

A fragment of the gene coding for residues 132–683 of human STAT1 was cloned in pET28, expressed in E. coli, and purified as previously described (54). Briefly, after ammonium sulfate precipitation, the protein was solubilized in a 20 mM Tris-HCl buffer at pH 8.8 containing 150 mM NaCl, and the solution was loaded onto an SP Sepharose column (GE Healthcare) equilibrated in the same buffer. The protein was not retained on the column and eluted in the flow through. After concentration, the protein was further purified by loading the solution containing STAT1_132–683 onto a Superdex S200 (GE Healthcare) size exclusion column equilibrated in buffer C.

Size exclusion chromatography-multiangle laser light scattering experiments

Size exclusion chromatography (SEC) combined with on-line detection by multiangle laser light scattering (MALLS) and refractometry is a method for measuring the absolute molecular mass of a particle in solution that is independent of its dimensions and shape (55). SEC was performed with a TSK G4000 (Tosoh Bioscience, Tokyo, Japan) or S200 Superdex column (GE Healthcare) equilibrated with 20 mM Tris-HCl buffer containing 150 mM NaCl, and the columns were calibrated with globular standard proteins of known hydrodynamic radius (R_h) (56). Chromatographic separations were performed at 20°C with a flow rate of 0.5 or 0.8 mL ⋅ min⁻¹, with on-line MALLS detection with a DAWN-HELEOS II detector (Wyatt Technology, Goleta, CA) using a laser emitting at 690 nm, and protein concentration was measured on line by the use of differential refractive index measurements using an Optilab T-rEX detector (Wyatt Technology) and a refractive index increment, dn/dc, of 0.185 mL ⋅ g⁻¹. The weight-averaged molecular mass was calculated using the ASTRA software (Wyatt Technology).

Small angle x-ray scattering experiments

Small angle x-ray scattering (SAXS) data were collected at the BioSAXS beamline (ID14-3) of the ESRF (57). The scattering from the buffer alone was measured before and after each sample measurement and used for background subtraction. All data analyses were performed using the program PRIMUS from the ATSAS 3.0.0 package (58).

NMR spectroscopy

The spectral assignment of the different constructs of NiV P protein (P_1–100, P_92–190, P_173–240, P_223–319, P_300–401, P_387–479, P_588–650, and P_XD (residues 655–709)) at 25°C in buffer B was determined using BEST-type triple resonance experiments (59). The NMR experiments were acquired at a ¹H frequency of 600 or 800 MHz (Agilent spectrometers). Six experiments were acquired for each construct: HNCO, intraresidue HN(CA)CO, HN(CO)CA, intraresidue HNCA, HN(COCA)CB, and intraresidue HN(CA)CB. All spectra were processed in NMRPipe (60) and analyzed in Sparky (61), and automatic assignment of spin systems was done using MARS (62), followed by manual verification. The assigned chemical shifts have been deposited in the BMRB database under accession numbers 50098, 50099, 50100, 50101, 50102, 50103, and 50105. The completeness of the assignments was above 97% for all constructs with the unassigned residues mostly belonging to the N-termini that undergo fast solvent exchange. Only the construct P_387–479 had a lower assignment completeness (92%) with residues S399-I402 being unassigned. Secondary structure propensities (SSPs) were calculated from the experimental C_α and C_β chemical shifts according to Marsh et al. (63).

X-ray crystallography

Initial crystallization conditions for the tetramerization domain (aa 471–580) were identified at the High Throughput Crystallization platform of the Partnership for Structural Biology (https://htxlab.embl.fr). The crystals used for data collection were obtained in 7% PEG 3350, 1 M LiCl, 0.1 M sodium citrate, 0.1 M arginine, 4% hexanediol, and 0.001% NaN3 at pH 5.0 and were frozen with 15% glycerol as cryoprotectant. X-ray diffraction data were collected at the ID29 beamline of the ESRF at a wavelength of 0.9793 Å and at the Proxima1 beamline of SOLEIL at a wavelength of 0.97895 Å and a temperature of 100 K and were processed with the XDS package (64). Initial phases were obtained using the anomalous scattering from selenium atoms by the SAD method with the program HKL2MAP (65). A model was initially constructed with the Autobuild program (66) from the phenix suite (67) and subsequently refined with the phenix.refine program (68) and Coot (69). The geometry of the final model was checked with MolProbity (70). In the model, 98.04% of residues had backbone dihedral angles in the favored region of the Ramachandran plot, 1.83% fall in the allowed regions, and 0.13% were outliers. Figures were generated with PYMOL (71) and Chimera (72).

Modeling of NiV phosphoprotein as an ensemble of conformers

An ensemble of 10,000 full-length NiV phosphoprotein conformers was generated with the program Flexible-Meccano (73). These included the high-resolution three-dimensional structures of the tetramerization domain (P_MD) and a homology model of the C-terminal X domain based on the known HeV P_XD structure (74) by using Modeler (75). To accurately model the disordered N-terminal region, the ASTEROIDS selection algorithm (76) was first applied to each of the subconstructs of the phosphoprotein to obtain an ensemble of conformers of these regions in agreement with experimental ¹H^N, ¹⁵N, ¹³C_α, ¹³C_β, and ¹³C′ chemical shifts. Briefly, in successive rounds of selection, 5 × 200 pairs of (ϕ, ψ) dihedral angles were selected for each residue from a pool of 10,000 conformers assembled by Flexible-Meccano such that the difference between ensemble-averaged ¹H, ¹⁵N, and ¹³C chemical shifts predicted by the program SPARTA (77) and corresponding experimental values was minimized. The selected dihedral angles were then used in generation of a new set of 10,000 conformers for the next round of selection. The (ϕ, ψ) databases for each subconstruct of the phosphoprotein thus obtained after four rounds of selection were combined and used by Flexible-Meccano to build conformations of full-length protein. Residues in the linker region between the tetramerization domain and the X domain were stochastically assigned amino-acid-specific backbone dihedral angles in agreement with random coil statistics. In the disordered regions, side chains were constructed using the program SCCOMP (78). Scattering curves were calculated with CRYSOL (ATSAS 3.0.0) (79) and ensemble selections were performed using ASTEROIDS (80) by combining both the experimental NMR and SAXS data. To account for conformational heterogeneity within different regions of the full-length protein, the ASTEROIDS target function combined SAXS data from three constructs: P_FL, P_1–580, and P_471–709, representing the full-length tetramer and the tetramerization domain with the N-terminal and C-terminal parts of the protein, respectively.

ITC

ITC experiments for characterizing the interaction between the five fragments P_92–190 (WT, Y116A, G120E, G125E, H117A/Q118A/H1119A) and STAT1_132–683 were performed with a MicroCal iTC200 system (Northampton, MA). The STAT1 solution at a concentration of 100 μM in buffer C was loaded in the calorimetric cell and titrated at 25°C, typically by performing 16 injections of 2.5 μL of the wild-type (WT) P_92–190 solution or of one of the variants (at concentrations ranging between 1.5 and 2.5 mM) in the same buffer. Protein samples and buffers were degassed before data collection. The heat generated for each injection was obtained by integrating the calorimetric signal. The heat resulting from the binding reaction between the two proteins was obtained as the difference between the heat of the reaction and the corresponding heat of dilution. The data were processed using Origin 7.0 (Malvern Panalytical, Malvern, UK) and fitted to a single-site binding model.

STAT1 localization in NiV-infected cells

Generation of recombinant NiV was described elsewhere (6). NiV-G121E and NiV-G135E contain a single amino-acid change in the P protein at position 121 or 135. To monitor the effect of NiV infection on endogenous STAT1, Vero E6 cells were grown on polymer coverslips of the μ-Slide VI^0.4 and were mock infected or infected with recombinant NiVs (WT, G121E, or G135E) at a multiplicity of infection of 0.5. At 1 h postinfection, the virus inoculum was replaced with Dulbecco’s modified Eagle’s medium containing 2.5% fetal bovine serum, and the cells were incubated at 37°C for 28 h. The cells were washed with serum-free medium and then treated with 1000 U/mL of human interferon-β in serum-free Dulbecco’s modified Eagle’s medium for 40 min at 37°C. For untreated controls, infected cells were incubated with IFN-free medium. Cells were washed twice with cold phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde-PBS solution for 20 min at room temperature, and subsequently treated with 0.1 M glycine for 10 min. Cells were permeabilized with 0.2% Triton X-100 in PBS for 10 min and washed with a blocking solution (2% bovine serum albumin (BSA) in PBS). Immunofluorescence staining was performed with mouse anti-phospho-STAT1 (pY701) (BD Biosciences, Franklin Lakes, NJ) (1:250 dilution) antibody or rabbit anti-NiV matrix protein (1:500) antibody.

Results

NiV phosphoprotein has a modular architecture

The presence of IDRs in a protein can be predicted from its amino-acid sequence by different computational approaches (81,82). Previously, we devised a metaprediction method that calculates a consensus disordered score (D-score) between 16 independent disorder predictors by a simple voting procedure (32) and found that by applying a threshold at 0.5 to distinguish between ordered and disordered regions, we were able to precisely predict boundaries between folded domains and IDRs (32,50, 51, 52), as well as the existence of MoREs (36).

The D-score profile calculated from NiV phosphoprotein sequence reveals three disordered regions of D-score < 0.5 (named IDR1–3) and several regions of D-score > 0.5 predictive of structural order (Fig. 1, A and B), in agreement with the hydrophobic cluster analysis plot (Fig. S1; (83)). Two of these structured regions of more than 100 aa between residues 475 and 598 (averaged D-score of 0.82) and between residues 658 and 709 (averaged D-score of 0.69) coincide with the known localization of the folded tetramerization (P_MD) and X domains (P_XD), respectively (37,39). A shorter region of D-score > 0.5 at the N-terminal end (aa 1–29) corresponds to the α-MoRE known to bind the unassembled nucleoprotein N⁰ (9). However, the D-score also identifies other putatively ordered regions. A region of D-score > 0.5, also spanning more than 100 aa between residues 102 and 219, encompasses the putative binding sites for STAT1, STAT2, and Polo-like kinase 1, as well as the CRM1-dependent nuclear export signal (Fig. 1 B). A shorter region of D-score >0.5 is localized within the C-terminal linker (aa 607–658). Finally, borderline regions appear between residues 310 and 450 in the form of clearly visible peaks in the D-score even if the score remains under the threshold of 0.5. One of these peaks encompasses the conserved ultraweak nucleoprotein binding motif recently discovered in measles virus phosphoprotein (10), whereas two other ones overlap with the regions encoding the specific moiety of the V and W proteins (Fig. 1 B).

Protein production and quality control

We expressed the full-length NiV phosphoprotein and its different fragments as recombinant proteins in E. coli (Fig. 1 C). The full-length protein and the different protein fragments were soluble and eluted as single peaks from the SEC column. The molecular mass determined by on-line MALLS detection was constant throughout the chromatographic peak (polydispersity index, M_w/M_n lower than 1.01), indicating that each sample was monodisperse (Table 1).

Table 1.

Mass and Size of the Different Phosphoprotein Constructs

Protein	Naaa	pIcalcb	MMcalc (kDa)c	MMexp (kDa)d	Rh,folded (nm)e	Rh,unfolded (nm)e	Rh,exp (nm)f	Rg,theo(Bernado) (nm)g	Rg,theo(Fitskee) (nm)h	Rg,exp nm)i	Dmax (nm)j
P1–100	108	5.1	12.347	13.5 ± 1.5	1.8	3.1	2.6 ± 0.2	2.9	3.4	3.2 ± 0.1	12
P92–190	108	4.9	11.875	12.0 ± 0.5	1.8	3.0	2.6 ± 0.2	2.9	3.4	3.1 ± 0.2	12
P173–240	78	6.4	8.691	8.0 ± 2	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
P92–250	168	4.6	18.583	21.0 ± 2.0	n.d.	n.d.	n.d.	3.7	4.5	4.0 ± 0.3	15
P223–319	106	4.5	11.536	12.9 ± 1.5	1.8	3.0	3.1 ± 0.2	2.9	3.4	3.1 ± 0.1	11.5
P300–401	111	5.9	12.421	11.5 ± 1.5	n.d.	n.d.	n.d.	2.9	3.5	n.d.	n.d.
P387–479	102	7.1	11.151	12.4 ± 1.5	1.8	3.0	2.5 ± 0.2	3.3	3.3	3.2 ± 0.1	12.5
P471–580∗	119	5.2	13.757 (55.028)	58 ± 6	3.1	6.7	4.3 ± 0.2	n.r.	n.r.	n.d.	n.d.
P655–709	63	5.3	7.364	8 ± 1	n.d.	n.d.	n.d.	2.2	2.5	1.6 ± 0.1	5.5
P588–650	72	5.3	8.289	8 ± 2	n.d.	n.d.	n.d.	2.4	2.7	2.6 ± 0.1	10
P588–709	131	4.9	15.041	15.3 ± 1	1.6	2.5	1.9 ± 0.2	n.d.	n.d.	n.d.	n.d.
P1–580∗	588	4.7	64.756 (259.024)	250 ± 25	5.5	14.8	9.2 ± 0.3	n.r.	n.r.	10.8 ± 0.3	41
P471–709∗	248	4.9	28.369 (113.476)	113 ± 10	4.1	9.8	7.8 ± 0.3	n.r.	n.r.	6.3 ± 0.1	22
PFL∗	717	4.7	79.367 (317.468)	306 ± 30	6.0	16.8	9.3 ± 0.2	n.r.	n.r.	10.9 ± 0.1	39

n.d., not determined; n.r., not relevant.

^∗Proteins that were found to be tetrameric in solution.

^aNumber of amino acids in the used constructs, including C-terminal His tag and linker (LEHHHHHH) and, in some cases, an N-terminal M.

^bIsoelectric point of the used constructs calculated using the webserver ProtParam. In parentheses, isoelectric point calculated without the C-terminal His tag.

^cIn parentheses, calculated mass for a tetramer.

^dMeasured by SEC-MALLS.

^eThe theoretical R_h-values are calculated from R_h-values of folded and chemically unfolded standard proteins taken from Uversky et al. (56) and using the experimental molecular mass determined by SEC-MALLS.

^fMeasured from SEC data by calibrating the system with standard globular proteins.

^gCalculated using the power law and parameters defined by Bernado et al. (88)

^hCalculated using the power law and parameters defined by Fitskee et al (87).

ⁱMeasured from SAXS data at low Q using the Guinier approximation.

^jDetermined from SAXS data using the program PRIMUS.

NiV full-length phosphoprotein is a tetramer of large size

The molecular mass of full-length NiV phosphoprotein (P_FL) measured by on-line MALLS is 306 ± 30 kDa, in agreement with the molecular mass calculated for the tetramer (4 × 79.367 = 317.468 kDa) (Fig. 2; Table 1). Its hydrodynamic radius (R_h) measured by SEC is 9.3 ± 0.2 nm, a value that is intermediate between those expected for a compact globular protein (R_h,theo = 6.0 nm) or an unfolded chain (R_h,theo = 16.8 nm) of the same molecular mass (Table 1). Analyses of the different protein fragments by SEC-MALLS indicate that the central region is solely responsible for the tetramerization of the protein, in accordance with cross-linking experiments (84) and the recent crystal structure of the NiV phosphoprotein multimerization domain (37). The fragments P_1–580 and P_471–709, which both encompass the tetramerization domain, and the fragment P_471–580, which closely corresponds to this domain_, have molecular masses determined by SEC-MALLS of 250 ± 25, 113 ± 10, and 58 ± 6 kDa (Fig. 2), respectively, in agreement with the theoretical molecular masses expected for their tetrameric forms (Table 1). Similarly to the full-length protein, the hydrodynamic radius of these fragments is intermediate between those expected for compact globular proteins and for unfolded chains of the same molecular mass (Table 1). By contrast, all other fragments are monomeric, and those from the N-terminal region have a hydrodynamic radius close to that of an unfolded chain (Table 1 and see below).

Figure 2.

Molecular masses and size of full-length phosphoprotein (P_FL in black) and of different fragments (P_1–580 in red; P_471–709 in green; P_471–580 in blue) in solution. SEC was performed at 20°C on a TSK G4000 column equilibrated with 20 mM Tris-HCl (pH 7.5) containing 150 mM NaCl using a flow rate of 0.8 mL ⋅ min⁻¹. The elution was monitored on line by MALLS and differential refractometry. The lines show the chromatograms monitored by differential refractive index measurements. The crosses indicate the molecular mass calculated from static light scattering across the elution peak, and the numbers indicate the weight-averaged molecular mass (kDa) with standard deviations. The hydrodynamic radius (R_h) was determined from the elution volume by calibrating the column with standard globular proteins (Table 1). To see this figure in color, go online.

SAXS curves were recorded at three different protein concentrations for the full-length phosphoprotein (Fig. S2) and for both tetrameric constructs: P_1–580 (Fig. S3) and P_471–709 (Fig. S4). For each protein, the scattering profile shows an absence of concentration dependence, and the Guinier plots at low Q values are linear, indicating the absence of aggregation or of significant interchain interactions (Figs. S2–S4). We note, however, an unexplained overestimation of more than 30% in the molecular mass of full-length phosphoprotein and of P_471–709 calculated by using the calibration with BSA (Table 2). The dimensionless Kratky plot and the pair-distance distribution function revealed ambivalent features for the three proteins, supporting the presence of both structured and disordered parts (Table 2; Figs. S2–S4; (85)). A peak slightly above 1.1 is visible in the Kratky plot at Q × R_g ∼ 2, likely reflecting the presence of the folded tetramerization and X domains. For full-length phosphoprotein and the fragment P_1–580, the value evolves toward a plateau at higher Q × R_g, indicating the presence of a large fraction of disordered regions in agreement with the asymmetric bell-shape curve in the P(r) plots. For the fragment P_471–709, a maximum is also present at Q × Rg ∼ 5 in the Kratky plot and two maxima at r ∼ 4 nm and r ∼ 8.5 nm, likely reflecting on the long-range contacts between the tetramerization domain and the four X domains. The radius of gyration determined from the Guinier plot and the longest pair distance D_max were almost identical for the full-length protein and the N-terminal fragment P_1–580, indicating the overall size of these molecules is mainly determined by the N-terminal disordered region (Table 1). The corresponding values for the C-terminal fragment P_471–709 were clearly smaller, in agreement with the largest fraction of residues involved in structured domains (∼65% of folded residues) (Table 1).

Table 2.

Data Collection and Structure Statistics for SAXS

	PFL	P1–580	P471–709
Data Collection Parameters
Instrument	ID14 with PILATUS detector (ESRF, Grenoble, France)
Beam geometry (mm)	0.7 × 0.7
Sample distance (m)	2.43
Wavelength (Å)	0.931
Q range (nm−1)	0.0406–5.0000
Exposure time	10 × 10 s
Temperature (K)	277
Concentration range (mg ⋅ mL−1)	1.0–1.8	0.5–1.0	3.5–5.8
Structural Parameters
Rg (nm) (Guinier)	10.9 ± 0.1	10.8 ± 0.3	6.3 ± 0.1
Rg (nm) (P(r))	11.2	11.3	6.4
Dmax (nm)	39	41	22
Molecular Mass Determination (kDa)
From BSA calibration	432	240	164
Calculated from sequence	317	259	113
Software Employed
Primary data reduction	BsxCuBE
Data processing	PRIMUS (ATSAS 3.0.0)
Ensemble modeling	ASTEROIDS
Three-dimensional representations	PYMOL (1.8.6)

Crystal structure of the tetramerization domain

The fragment P_471–580 crystallized in different conditions and different space groups. We solved the structure by the SAD method using selenomethionine-labeled protein crystallized in space group P12₁1 (see Table 3 for details of data collection and statistics) The asymmetric unit contains two copies of a tetramer (Fig. S5, A and B), in which each monomer is composed of two short helices (helices α₁ and α₂) and a long C-terminal helix (helix α₃: aa 508–570) (Fig. 3, A and B). The long helix of each monomer is composed of eight complete heptad repeats and a stammer at position 540–542, resulting in a kink near residue Pro544 (Fig. 3 C). Within one tetramer, the four long helices fold into a parallel coiled coil forming a central shaft, whereas the short helices form a crown around the N-terminal end of this shaft (Fig. 3, A and B). The global net charge of the tetrameric domain is negative at neutral pH, and the calculated electrostatic surface potential shows an asymmetrical distribution with a negative head and a patch of negative charges at the C-terminal end of the coiled-coil shaft, whereas the central part of the coiled-coil shaft is rather neutral (Fig. S5 C).

Table 3.

Crystallographic Data Collection and Refinement Statistics

Parameter	Value
Resolution (Å)	48.75–3.00 (3.06–3.00)
Space group	P1 21 1
Unit cell
a, b, c (Å)	59.5, 85.60, 122.6
α, β, γ (°)	90.0, 94.7, 90.0
Total reflections	195,091
Unique reflectionsa	47,924
Rfree test set multiplicity	1257
	4.1 (1.9)
Completeness (%)	99.5 (99.9)
I/σ(I)	5.3 (2.27)
Rwork	0.175 (0.253)
Rfree	0.229 (0.286)
RMSD (bond lengths) (Å)	0.013
RMSD (angles) (°)	1.16
Ramachandran favored (%)	98.04
Ramachandran allowed (%)	1.83
Ramachandran outliers (%)	0.13
Average B factor (Å2)	55

Values in parentheses are for the highest- resolution shell.

^aFriedel’s pairs unmerged.

Figure 3.

Crystal structure of NiV phosphoprotein tetramerization domain (P_MD). (A) A cartoon representation of tetrameric NiV P_MD (PDB: 4GJW) is given. The different chains are shown in different colors, and the overall dimensions are indicated (black arrows). The N- and C-terminal ends are indicated for each chain, and the three α-helices, α₁ to α₃, are shown for chain D (in red). The position of the stammer in the long coiled coil is shown, and the dotted lines indicate the location of the a-layers displayed in (D) and (E). (B) Contacts between the chains within the crown are shown. A cartoon representation of the NiV tetramerization domain showing that each chain contacts the other three chains is given. (C) The sequence alignment of the tetramerization domain of members of the genus Henipavirus is shown. (HeV: Hendra virus; CeV: Cedar virus; GBHV: Ghanaian bat henipavirus). The letters under the sequences indicate the position of the different heptad repeats (h_1–8) and of the stammer. A and D positions in the heptad repeats, which correspond to residues pointing inside the coiled coil, are shown in green to indicate hydrophobic residues or in blue to indicate polar residues. (D and E) Symmetry of layers in the coiled coil is shown. Stick representations of two A-layer amino acids (Cys512 and Leu557) showing the transition from fourfold to twofold symmetry along the coiled coil (see positions in A) are given. (F) Sequence conservation map is shown. The space-filling model of the NiV tetramerization domain, in which surface residues are colored according to the degree of conservation (using the multiple sequence alignment of C), is shown. Figures were drawn with PYMOL. To see this figure in color, go online.

The N-terminal part of the tetrameric domain exhibits fourfold symmetry, but in the coiled-coil shaft, the fourfold symmetry is progressively replaced by a twofold symmetry when going from the N-terminal end to the C-terminal end (Fig. 3, D and E). The monomers within a tetramer superimpose by pairs with C_α root mean-square deviation (RMSD) values around 0.6 (twofold symmetrical mate) and 1.0 Å (other two chains). Each protomer N_i contacts the three other protomers. The long helix α₃ of N_i is packed between helix α₃ of the N_{i − 1} and N_{i + 1} protomers, whereas helix α₂ contacts the N_{i + 1} protomer and helix α₁ contacts the N_{i + 2} protomer (Fig. 3 B). In the corresponding tetrameric domain of Sendai virus phosphoprotein, each protomer only interacts with two other adjacent protomers (38).

The tetrameric structure of NiV tetramerization domain is stabilized by multiple interchain interactions, each protomer forming on average 19 H-bonds and 11 salt bridges with the others protomers and burying a surface area of 2200 Å² between each pair of adjacent protomers (for a total of 8800 Å² of accessible surface area buried in the tetrameric structure). The interior of the coiled-coil shaft is composed of 19 side-chain layers, mainly hydrophobic but with four layers of polar residues (C512, S515, G519, and A522) at the N-terminal end and one near the C-terminal end (T560) (Fig. 3 C).

The structure is very similar to the crystal structure of NiV tetramerization domain published by Saphire and co-workers (37), with, however, two noticeable differences: 1) although the constructs are similar in length, the coiled-coil region in our structure is six residues shorter than in the previous structure (62 aa instead of 68 aa), with a length intermediate between that of Sendai virus (64 aa) (38) and that of measles virus (60 aa) (86), suggesting crystal contacts play a role in the stabilization of this part of the structure. 2) The C-terminal part of the coiled coil exhibits a slightly different superhelical frequency (−3.39° vs. −3.63°), resulting in a different superhelical pitch (158 Å vs. 148 Å) (Fig. S5 D). The mapping of crystallographic B-factors onto the structure confirms that the C-terminal end of the coiled-coil shaft is more flexible than the remaining part of the molecule (Fig. S5 E), as shown previously (34,37).

The N-terminal region and the C-terminal linker are globally disordered

Several fragments of ∼100 amino acids covering the entire N-terminal region but also the linker in the C-terminal part (P_CT-LINK), which is predicted to be at least partially disordered (IDR3 in Fig. 1 A), and the X domain were produced as ¹³C, ¹⁵N-labeled proteins and purified (Fig. 1 C). Each fragment from the N-terminal region has a hydrodynamic radius (R_h) significantly larger than that expected for a globular protein of the same molecular mass and slightly smaller than that expected for a fully unfolded protein in a good solvent (Table 1). SAXS data recorded for these fragments revealed no concentration dependence. For the different N-terminal fragments and for the fragment corresponding to the C-terminal linker (P_588–650), the experimental radius of gyration (R_g) values determined using the Guinier approximation were close to the R_g-values calculated for a random coil according to Flory’s power-law dependence on chain length using parameters for chemically unfolded proteins (87) or for intrinsically disordered proteins (Table 1; (88)), in line with the proposal that these regions are disordered. By contrast, the radius of gyration for the X domain is significantly smaller than that of a disordered chain of the same length, as expected for a folded X domain similar to that of Hendra virus (74).

To characterize the structure of these fragments in solution, we carried out assignment of the backbone NMR resonances and calculated their SSPs from the experimental C_α and C_β chemical shifts. The ¹H-¹⁵N two-dimensional heteronuclear single quantum coherence (HSQC) spectra of the different N-terminal fragments exhibit poor chemical shift dispersion of the amide ¹H resonances as is typical for a disordered protein (Fig. S6; (89)). The combined SSP score for the entire N-terminal region (P_NTR) reveals the presence of several transient α-helices (Fig. 4 A). In particular, we observe two transiently populated helical elements at the very N-terminus that make up the binding region for N⁰ (9). In the N⁰-P complex, these helical elements (aa 1–29) bind to the C-terminal domain of the nucleoprotein mainly through hydrophobic interactions, forming a stable helix kinked at residue Q21 (9). In addition, a third helical element with a population of around 40% is observed at residues 340–355, which coincides with the putative additional weak binding region for the nucleoprotein recently discovered in measles virus (10). The ¹H-¹⁵N HSQC spectrum of the fragment P_588–650 corresponding to the C-terminal linker (P_CT-LINK) is also typical of a disordered protein (Fig. S6), and the combined SSP score also revealed the presence of low-populated α-helices (aa 600–625) (Fig. 4 B). By contrast, the ¹H-¹⁵N HSQC spectrum of the isolated X domain, (P_XD, aa 655–709) confirms that this domain is folded in solution, and the chemical shift assignments clearly show that a three-helix bundle structure is predominant in solution (Fig. 4 C), as also observed in the corresponding domains of measles, mumps, Sendai, and Hendra viruses (74,90, 91, 92).

Figure 4.

NMR characterization of NiV phosphoprotein N-terminal region (P_NTR), C-terminal linker, and X domain. (A) Secondary structure propensities (SSPs) of the N-terminal region of the phosphoprotein obtained on the basis of experimental ¹³C_α and ¹³C_β chemical shifts are shown. The propensities were obtained by combining spectral assignments of individual subconstructs of this region. (B) SSPs of the C-terminal intrinsically disordered linker region of the phosphoprotein (P_CT-LINK) obtained from experimental ¹³C_α and ¹³C_β chemical shifts are shown. (C) SSPs of the C-terminal X domain of the phosphoprotein (P_XD) obtained from experimental ¹³C_α and ¹³C_β chemical shifts are shown. (D) A ¹H-¹⁵N HSQC spectrum of the full-length NiV phosphoprotein (red) superimposed with HSQC spectra (blue) of individual subconstructs of the protein (P_1–100, P_92–190, P_173–240, P_223–319, P_300–401, P_387–479) is shown. To see this figure in color, go online.

The HSQC spectrum of the tetrameric full-length phosphoprotein (a tetrameric protein complex of 2868 aa with our construct) contains resonances with limited chemical shift dispersion that superposed well with the resonances from the residues of the N-terminal disordered part of the protein (Fig. 4 D; Fig. S6). This clearly indicates that the N-terminal part is flexible, with similar conformational behavior in the tetramer as in the isolated peptides and negligible interchain contacts. No resonances for the tetrameric domain, C-terminal linker, or the X domain could be identified in the spectrum of full-length phosphoprotein, possibly reflecting slower rotational tumbling caused by a hydrodynamic drag because of the connection to the tetrameric domain and the presence of the X domain at the C-terminus, although we cannot rule out the possibility of interactions between the four C-terminal regions that may also give rise to broadening.

Overall, these results show that 1) the N-terminal region of NiV phosphoprotein is intrinsically disordered and exchanges between multiple conformers on a fast timescale, 2) this region contains transient α-helical elements, 3) there are no significant interactions between the N-terminal region of the four protomers within the tetrameric phosphoprotein, and 4) the C-terminal linker also seems to be disordered and to contain transient α-helical elements.

Ensemble description of full-length NiV phosphoprotein tetramer

To study the conformational diversity of NiV phosphoprotein, we generated ensembles of all-atom conformations of the full-length protein and of both tetrameric fragments (P_1–580 and P_471–709). We used our crystal structure of the tetrameric domain (Fig. 3), a homology model of the X domain based on the known structure of the X domain of Hendra virus (74), heteronuclear backbone chemical shifts (C_α, C_β, C′, and ¹⁵N) for the N-terminal fragments (Fig. 4), and SAXS data for the three proteins (Figs. S2–S4) as experimental constraints to delineate the contours of conformational space sampled by the full-length protein. We used a combine fit of multiple curves to select conformational subensembles of 10 conformers for which the three average theoretical scattering curves reproduced the three experimental curves. The average SAXS curve calculated for this ensemble is comparable to the experimental SAXS curve measured for the full-length phosphoprotein, whereas the average curves calculated with the properly truncated forms of the same 10 conformers are comparable to the experimental SAXS curves obtained for P_1–580 and P_471–709 (Fig. 5 A). A global χ²-value of 1.53 and the plot of reduced residuals indicate the quality of the fit (Fig. 5 A). It should be noted that the independent fits performed for each of the three proteins yielded ensembles of conformers with averaged R_g-values in agreement with the experimental values (full-length P: 13 ± 3 nm, P_471–709: 6.3 ± 0.5 nm, P_1–508: 12 ± 3 nm) but did not improve the overall quality of the fits (Fig. S7). Some discrepancies between the experimental and fitted curves are visible. The downward deviation observed at low Q-values for full-length phosphoprotein and to a lesser extent for the P_1–580 fragment could be due to intermolecular interactions, whereas the discrepancy observed between the experimental and calculated curves for the fragment P_471–709 at Q-values > 1.0 nm⁻¹ (Fig. 5 A) might result from nonoptimal buffer subtraction, as already observed for proteins such as this fragment, which comprises both folded and unfolded domains that may have distinct electron densities (93,94). Alternatively, it can also result from a bias in the construction of the initial ensemble of the full-length protein. Because the initial ensemble of conformers was built for full-length protein and was then truncated to generate the initial ensembles for the fragments P_1–580 and P_471–709, it is possible that the large N-terminal disordered region has constrained the X domains to position at one pole of the tetramerization domain, whereas such constraints are absent in the fragment P_471–709. Finally, additional structural organization or interchain interactions might be present in the C-terminal part of the phosphoprotein, which has not been taken into account in our models. It is thus remarkable that the same set of conformers is able to represent the curves obtained for the three constructions. The R_g-values of the initial ensemble for the full-length protein exhibited broad Gaussian distribution centered around 12 nm, and the R_g-values of the selected ensembles of 10 conformers exhibit a similar broad distribution (Fig. 5 B), suggesting the absence of constraints on the overall size of the protein. Fig. 5 C shows a typical model of the full-length phosphoprotein and Fig. 5 D shows the overlays of the 10 conformers of each construct selected in the same fitting process for reproducing the data shown Fig. 5 A.

Figure 5.

Ensemble structure of the NiV phosphoprotein. (A) Multiple-curve fitting of SAXS data is shown. The upper panel shows experimental SAXS curves of the three tetrameric proteins: P_FL (in red), P_1–580 (in blue), and P_471–709 (in orange), with experimental standard deviations shown as error bars. The black lines show the multiple-curve fit obtained for the three SAXS curves for a selected ensemble of 10 conformers. The reduced χ²-value is 1.53 over all three constructs. The lower panels show the reduced residuals (ΔI/σ). (B) Rg distribution for an ensemble of 10 conformers of full-length phosphoprotein is given. The thin black bars show the distribution for the initial ensemble of conformers. The large red bars show the R_g distribution for selected ensembles of 10 conformers. (C) Single model of NiV phosphoprotein structure is shown. The tetramerization domain and X domain are shown as surface representations. The different chains are shown in different colors, and the N-terminal end and X domain (P_XD) are indicated for each chain. (D) Ensemble representations of the full-length NiV phosphoprotein (P_FL) and of the two tetrameric fragments (P_1–580 and P_471–709) are shown. The 10 different models are aligned by superimposing their tetramerization domain. The tetramerization domain and X domain are shown as surface representations, and the different chains are shown in same colors as in C. The models of the P_471–709 fragment are shown in two orientations. To see this figure in color, go online.

The N-terminal region of NiV phosphoprotein binds human STAT1 with low affinity

The region of NiV phosphoprotein that binds STAT1 was previously mapped between residues 100 and 160 (43,47), which contains several highly conserved residues among henipaviruses (Fig. 6 A). Point mutations within residues 81–120 of the phosphoprotein affected the expression of a minireplicon, suggesting that this region is involved in the functioning of the polymerase machine, whereas mutations within residues 111 and 140 prevented the inhibition of interferon signaling (6).

Figure 6.

Interaction of the N-terminal region of the NiV phosphoprotein (V and W proteins) with STAT1. (A) Sequence conservation between residues 85 and 140 of the phosphoprotein (V and W proteins) across different henipaviruses is shown (see Fig. 3C). The labels above the alignment indicate the mutants used in the ITC experiments (in blue) and in the generation and assays of recombinant viruses (in green). (B) A schematic representation of STAT1 architecture is shown. The color boxes show the different domains present in our construct, which encompasses residues 135–683 of STAT1, whereas the gray boxes indicate the missing part. (C) SEC-MALLS analysis is given. The lines show the chromatograms monitored by refractometry, and the red crosses show the molecular mass calculated from on-line static light scattering measurements (MALLS). The upper panel shows the results for STAT1_132–683, the middle panel for the fragment P_92–190, and the lower panel for the mixture of the two protein fragments in a 10:1 ratio after 2 h incubation at 20°C. (D) Overlay of a region of the two-dimensional ¹H-¹⁵N HSQC NMR spectra of P_92–190 (∼540 μM) in the absence (in red) and presence of STAT1_132–683 (∼54 μM) (in blue) is shown. (E) Intensity profile was calculated as the ratio of peak intensities in the ¹H-¹⁵N HSQC spectrum of P_92–190 in isolation (I₀) and in the presence of STAT1_132–683 (I). To see this figure in color, go online.

To investigate the interaction between NiV phosphoprotein and STAT1, we used the fragment P_92–190 that contains the entire putative binding site for STAT1, and we expressed and purified a fragment of human STAT1 that comprises residues 132–683 encompassing the coiled-coil domain, the DNA binding domain, the linker domain, and the SH2 domain (Fig. 6 B; (95)). The missing 131 N-terminal residues and 38 C-terminal residues from the STAT1α isoform are not required for interaction with the phosphoprotein or V or W proteins of NiV (44). The protein STAT1_132–683 eluted from the SEC column as a single peak, and its molecular mass determined by SEC-MALLS of 64 ± 6 kDa is in good agreement with the theoretical mass of 64.1 kDa (Fig. 6 C). The fragment P_92–190 alone also eluted as a single peak, and its molecular mass of 11 ± 1 kDa is in agreement with the theoretical mass of the monomer. In a first approach, we mixed STAT1_132–683 (∼54 μM) with P_92–190 (∼540 μM) and loaded the mixture on a SEC column after 2 h incubation at 20°C. Both proteins eluted independently at elution volumes and with molecular masses corresponding to isolated STAT1_132–683 and P_92–190 (Fig. 6 C). In a second approach, we used NMR spectroscopy and recorded ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled P_92–190 at a final concentration of ∼22 μM in the absence or presence of unlabeled STAT1_132–683 (∼44 μM final concentration) (Fig. 6 D). In a complex of the size expected for the heterodimer STAT1_132–683–P_92–190 (∼75 kDa), NMR signals are strongly broadened in protonated samples, thus precluding their detection. By comparing the HSQC spectra, we find a partial reduction of the intensity in the presence of STAT1_132–683 for resonances corresponding to residues 100–150 (Fig. 6 E). The most affected resonance intensities are within the region of residues 110–140, which contains several residues of crucial importance for blocking STAT1 signaling in cell-based assays (6). We interpret this result as evidence that, under these experimental conditions, only a fraction of P_92–190 molecules bind STAT1_132–683, and we estimate the dissociation constant for STAT1_132–683 in the range of 10–100 μM. We note that conformational exchange contributions from the exchange between free and STAT1-bound P_92–190 could also be the source of some of the observed line broadening in the interaction experiments.

We used ITC and site-directed mutagenesis to further characterize the interaction between the two protein fragments. We used STAT1_132–683 as receptor at a concentration of 100 μM and performed titration with P_92–190 as ligand at an initial concentration between 1.5 and 2.5 mM. Despite that complete saturation was not achieved, the binding isotherm can be fitted to a single binding site model, yielding a dissociation constant (K_d) of 100 ± 15 μM ($\Delta G_{Bind}^0$ = +RTlnK_d = −5.4 kcal ⋅ mol⁻¹) (from five independent measurements) (Fig. 7 A), in good agreement with the NMR interaction experiment (Fig. 6 E). The favorable enthalpy of binding ($\Delta H_{Bind}^0$) of −9.5 kcal ⋅ mol⁻¹ is partially balanced by an unfavorable entropic contribution of 4.1 kcal ⋅ mol⁻¹ ($\Delta S_{Bind}^0$ = −14 cal ⋅ mol⁻¹ ⋅ K⁻¹) (Table 4).

Figure 7.

Effect of mutations on the interaction between the phosphoprotein and STAT1. Calorimetric titration of STAT1_132–683 with P_92–190 WT (A), G120E (B), G125E (C), Y116A (D), and the triple mutant (H117A/Q118A/H119A) (E) is shown. The upper panels show the heat effects associated with the injection of P_92–190 (2.5 μL per injection of ∼2 mM solution) into the calorimetric cell (200 μL) containing STAT1_132–683 at a concentration of 100 μM. All experiments were performed in 20 mM phosphate buffer (pH 7.5) containing 150 mM NaCl at 25°C. The lower panels show the binding isotherm corresponding to the data in the upper panels, and the lines show the best fitted curves for a 1:1 binding model. (F) Average dissociation constant obtained from three experiments for WT P_92–190 and the different variants is given. AAA = H117A/Q118A/H119A. Error bars represent standard deviation. (G) Impact of NiV infections (NiV WT, NiV-G121E, NiV-G135E) on the localization of phosphorylated endogenous STAT1 is shown. Vero E6 cells were infected with the indicated viruses and treated with interferon-β at 28 h postinfection. The phospho-Y701 form of endogenous STAT1 was visualized in paraformaldehyde-fixed cells 40 min after interferon-β addition (phospho-STAT1, green staining). NiV matrix (M) protein expression is a marker for infection (red). Arrowheads point to representative noninfected cells and nuclei stained for phospho-STAT1. Arrows indicate cells infected with NiV-G121E mutant defective in STAT1 binding and again nuclei stained for phospho-STAT1. To see this figure in color, go online.

Table 4.

ITC Data

Ligand	n	Kd (μM)	$\Delta {\text{G}}_{\text{Bind}}^0$ (kcal ⋅ mol−1)	$\Delta {\text{H}}_{\text{Bind}}^0$ (kcal ⋅ mol−1)	$-\text{T}\Delta {\text{S}}_{\text{Bind}}^0$ (kcal ⋅ mol−1)	$\Delta {\text{S}}_{\text{Bind}}^0$ (cal ⋅ mol−1 ⋅ K−1)
wt	1.19	100 ± 15	−5.4 ± 0.8	−9.5 ± 1.2	4.2 ± 1.2	−14 ± 4.0
Y116A	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
H117A/D118A/H119A	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
G120E	1.23	70 ± 9	−5.7 ± 0.7	−8.4 ± 0.6	2.7 ± 0.6	−8.9 ± 2.1
G125E	1.31	210 ± 90	−5.0 ± 2.0	−7.5 ± 2.6	2.5 ± 2.7	−8.4 ± 9.2

T = 298 K. n.d., not determined.

Point mutations Y116A and G125E, as well as a triple alanine mutation H117A/D118A/H119A (¹¹⁷AAA¹¹⁹), in full-length NiV phosphoprotein were previously shown to block interferon signaling and to disrupt the interaction with STAT1 (6,46). By contrast, the single amino-acid mutation G120E had no effect (6). We expressed and purified NiV phosphoprotein fragments encompassing residues 92–190 (P_92–190) and harboring these different mutations (Y116A, G120E, G125E, and H117A/Q118A/H1119A). Each protein eluted as a single peak from the SEC column with a molecular mass calculated from MALLS in agreement with the expected molecular mass for a monomeric protein. We performed calorimetric titrations of STAT1_132–683 with these different variants, but we were unable to detect any interaction with the variants Y116A and H117A/Q118A/H1119A. With the G120E variant, the affinity was comparable to that of the WT protein (Fig. 7, B–F; Table 4). With the G125E variant, the affinity is slightly lower than that of WT protein, with a K_d-value of 210 ± 90 μM, although the difference in $\Delta G_{Bind}^0$ of 0.4 kcal/mol is not significant (Table 4). These results confirm that the interaction between the N-terminal region of NiV phosphoprotein and STAT1 is direct but weak.

To further highlight the importance of the STAT1-binding domain (amino acids 110–140) that is present in the phosphoprotein but also in the V and W proteins, we also used recombinant WT or mutant viruses, which were generated by reverse genetics (6) and possess either an intact or a defective P gene. Two mutant viruses containing substitutions at positions 121 and 135 of the phosphoprotein and V and W proteins (NiV-G121E and NiV-G135E) were compared with NiV WT for their ability to block translocation into the nucleus of endogenous tyrosine-phosphorylated STAT1 activated by interferon (Fig. 7 G). NiV matrix protein was used as a marker of infection, and a specific anti-tyrosine-phosphorylated STAT1 antibody was used to monitor the localization of NiV P gene specific proteins (Fig. 7 G). Little or no tyrosine-phosphorylated STAT1 was detectable in the nucleus of cells infected with the WT NiV, whereas adjacent, uninfected cells exhibited clear nuclear phospho-STAT1 staining. By contrast, a loss of STAT1 binding ability in NiV G121E resulted in strong and apparent tyrosine-phosphorylated STAT1 signal present in the nuclei (Fig. 7 G). Interestingly, NiV-G135E mirrored NiV WT and showed a similar pattern of disruption of normal STAT1 trafficking. Earlier, it had been shown that the V protein harboring the same substitution failed to detectably bind STAT1 in immunoprecipitation studies but inhibited ISG54-driven reporter induction, albeit less efficiently (6). Our data suggest that this mutant retains sufficient STAT1-binding activity but that the interaction is too weak to be seen in simple coexpression coimmunoprecipitation assays. The mechanism for such behavior will need to be further investigated, in particular with respect to the implication of STAT2.

Discussion

The highly disordered, tetrameric NiV phosphoprotein is an integral component of the transcription/replication machinery of this dangerous virus and of all members of the Mononegavirales and a hitherto largely unexplored target for viral inhibition. Because of the size and flexible nature of the phosphoprotein, its conformational behavior has remained unexplored.

Here, we present the first, to our knowledge, atomistic representation of full-length NiV phosphoprotein in solution. We show that under our experimental conditions, full-length NiV phosphoprotein, as well as two other long fragments containing the tetramerization domain, also form tetramers in solution (Fig. 2), and we have obtained a second crystal structure of this tetramerization domain (37,84). The phosphoprotein of other paramyxoviruses also contains a tetrameric domain (38,86,91,96), although the existence of a trimeric form in solution has been described (96,97). NiV tetramerization domain has dimensions and crystal structure similar to those of Sendai virus tetramerization domain, with a parallel tetrameric coiled coil and a crown of helices at its N-terminal end (38). By contrast, the tetramerization domains of measles and mumps viruses lack the crown of helices, and although the structure of measles domain is also a parallel tetrameric coiled coil, that of mumps domain is composed of two pairs of parallel helices packed in opposite orientation (86,91). Phylogenetic trees based on the phosphoprotein sequences but also on the whole-genome sequences place the genus Henipavirus in an intermediate position between the genera Morbillivirus (measles virus) and Respirovirus (Sendai virus), whereas the genus Rubulavirus (mumps virus) is more distant (98,99). The crown of helices may thus be an ancestral feature lost by some viruses or a modern one acquired by others that stabilizes the central coiled-coil shaft or provides additional function. Up to now, no function has been specifically associated with the N-terminal part of the tetramerization domain, and mutations that perturb the crystallographic water molecules inside the coiled-coil shaft at the crown level had no effect on phosphoprotein function in a minireplicon system (34). Mapping the crystallographic B-factors on the domain structure shows that the N-terminal part of the central coiled coil, upstream of the kink, is more rigid than the downstream part (Fig. S5 E). Although the crystallographic packing can influence B-factors, the conservation of this dynamical feature in the tetramerization domain of the related measles virus and Sendai virus argues for functional relevance. The presence of interhelical networks of H-bonds and salt bridges in the crown and in the upper part of the coiled coil preceding the kink region could explain this additional rigidity of the structure. By contrast, a conserved basic patch in the C-terminal part of the shaft is known to interact with the polymerase and mutations in this part affect polymerase activity (Fig. 3, C and F; (34)). Multiple sequence alignment of Henipavirus phosphoprotein shows that the crown is not the most conserved part of the tetramerization domain, whereas it reveals a highly conserved region following the basic patch and straddling the C-terminal end of the coiled coil, and the N-terminal part of the P_CT-LINK (aa 565–600) (Fig. 3, E and F). This region includes the less rigid part of the structure (Fig. S5 E), and in related Mononegavirales, it interacts directly with L polymerase (16,17). Because the polymerase changes conformation while it operates and moves along its RNA template, the interaction between the phosphoprotein and the polymerase is submitted to tensional, shear, and/or rotational stresses and thus likely requires some flexibility in the protein interfaces to maintain the attachment.

Previous sequence analyses predicted that the N-terminal part of the phosphoprotein encompassing the region common to the V and W proteins and extending up to the tetramerization domain is an IDR except for the N-terminal end (31,100), which consists of an MoRE that binds and chaperones unassembled N⁰ (9). The D-score calculated as a consensus of multiple disorder predictions reveals a more complex picture with intervening putatively ordered regions and undefined regions (Fig. 1 A). Experimentally, we confirm the overall disorder of the N-terminal region of the phosphoprotein. The hydrodynamic radius determined by SEC and the radius of gyration determined by SAXS indicate that each fragment in the N-terminal region (Table 1) is rather extended, in accordance with the behavior of intrinsically disordered polypeptide chains in water (88). The poor chemical shift dispersion of the NMR ¹H-¹⁵N HSQC spectra also supports the high flexibility of the different fragments, but the assignment of the spectra revealed the presence of transient secondary structure elements (Fig. 4 D; Fig. S6), and an overlay of the SSP parameter (Fig. 4 A) and D-score (Fig. 1 A) plots shows that in several instances, short regions with an intermediate D-score (moderately above or below 0.5) are transiently folded into helical conformation and could correspond to an α-MoRE involved in recruiting different partners, in particular the polo-like kinase 1, the exportin, or the viral nucleoprotein (Fig. S8). The comparison of the NMR spectra from the different N-terminal fragments with the full-length phosphoprotein shows that most resonances overlap (Fig. 4 D), thereby showing that the structural properties of fragments are the same in the full-length protein. In other words, the overall disorder of N-terminal region is not a result of our divide-and-conquer approach but is an intrinsic property of this region of the protein. This view is also confirmed by the ensemble modeling, which shows that an ensemble model of full-length phosphoprotein, in which the N-terminal region and the C-terminal linker are highly flexible and adopt random coil conformation, is in agreement with the SAXS and NMR data (Fig. 5).

The ensemble model of full-length phosphoprotein provides new, to our knowledge, insights about the conformation of this protein in solution (Fig. 5). Overall, the molecule appears elongated in the direction of the tetramerization domain shaft, with the four X domains clustering at the C-terminal end (Fig. 5 D). This spatial arrangement certainly provides an ideal positioning of the four X domains for interacting with the polymerase or the nucleocapsid and for “walking” along the latter, driven by the polymerase. The four N-terminal chains mainly occupy the other end and the sides of the tetramerization domain, likely creating the steric hindrance that pushes the X domains toward one end. When multiple decoys are overlaid with respect to the tetramerization domain, thus mimicking the time average spatial occupancy of the four flexible N-terminal chains and four C-terminal linkers (Fig. 5 D), the central tetramerization domain, including the basic site of putative interaction with the L polymerase, is masked. The recent structure of the polymerase complexes of two related viruses revealed that one of the chains of the phosphoprotein is mainly involved in the interaction with the L polymerase, but the three other chains are also binding (16,17). These multiple interactions could be required to move the N-terminal chains of the phosphoprotein away from the multimerization domain and to allow access of the polymerase to the central tetrameric shaft.

The tetrameric NiV phosphoprotein is thus a large molecule, with a hydrodynamic radius of 9.3 ± 0.3 nm and a radius of gyration of 10.9 ± 0.2 nm. The R_g/R_h ratio is a sensitive indicator of chain compactness; experimental values for this ratio lie between limits of ∼0.8 for globular proteins (101) and of ∼1.5 or higher for random coils (102). The value of 1.18 found here for NiV full-length phosphoprotein is in agreement with an elongated molecule and/or the presence of long IDRs and with the values of 1.0 and 1.14 that we previously reported for the full-length phosphoproteins of vesicular stomatitis virus and rabies virus (103), respectively. Although these phosphoproteins of the family Rhabdoviridae are dimeric and less than half its size, they share a similar architecture with the NiV phosphoprotein (32). Assuming a partial specific volume of 0.73 cm³ ⋅ g⁻¹, the intrinsic molecular volume of a protein the size of tetrameric NiV phosphoprotein should be ∼390 nm³, whereas the volume of the equivalent sphere calculated by using the experimental R_h-value is 3400 nm³, almost 10-fold larger. Multiple copies of this protein are packaged in the virion and might occupy an important part of the particle volume if not otherwise compacted. With respect to the nucleocapsid, the length of the disordered C-terminal linker (70 aa) allows an extension of the order of 50 nm (2 × 70 × 0.35 nm) between two X domains, which would allow to connect nucleoproteins at least 10 subunits apart along the nucleocapsid, and probably even more distant if the possible extension of the C-terminal disordered region of the nucleoprotein (N_TAIL) is taken into account. In the virion, the nucleocapsids have a diameter of 18–19 nm and an average pitch of 5 nm (7), and thus, one phosphoprotein could connect nucleoproteins across the nucleocapsid. The larger size of tetrameric NiV phosphoprotein may thus be of importance in the packaging of the viral particle and in the functioning of the polymerase machine, catching at long distance and directing partners to their site of use.

The architecture of the phosphoprotein is conserved among Mononegavirales, including the localization of functional and structural modules, but the length of the protein varies considerably between families and even between genera within the same family. As illustrative examples, VSV in the family Rhabdoviridae is 267 aa long, whereas measles virus and NiV phosphoproteins in the family Paramyxoviridae are 507 and 709 aa long, respectively. The reason why the NiV phosphoprotein is so long remains enigmatic, considering that the VSV phosphoprotein, for example, perfectly assumes similar functions in the life cycle of the virus. In a previous study, we compared the structure of the C-terminal N-binding domain of the phosphoprotein from different Mononegavirales and discovered that the variation in length of this protein domain correlated with the capacity to perform additional functions, suggesting a model in which the protein evolves by accretion of functional modules (52). Henipavirus phosphoproteins are significantly longer than the phosphoproteins from other genera of Paramyxoviridae (104). Because the L polymerases of the Mononegavirales have a high error rate, the genome of these viruses has the capacity to rapidly evolve and trim superfluous parts of the genome, suggesting that the very existence of additional lengths in the NiV phosphoprotein sequence is a sign that they perform additional functions or are required for the fitness of the virus. We can therefore speculate that the larger size of the N-terminal domain correlates with its implication in other functions and that the transient α-helical elements evidenced by NMR spectroscopy and predicted by sequence analysis represent short linear motifs that recruit different and yet unidentified partners. Some of them clearly encompass regions of the protein known to bind viral or cellular proteins (Fig. S8). A comparison with the measles virus phosphoprotein shows that most functional and structural regions are conserved but that the NiV phosphoprotein contains an additional region in the N-terminal region between residues 150 and 300 that is common to the V and W proteins and is required for binding STAT2 (Fig. S9). The binding site of STAT1 appears to be conserved between NiV and measles virus, being localized between residues 110 and 140, but in measles virus, the binding of STAT2 is mediated by the specific C-terminal region of the V protein (105). We speculate that the NiV phosphoprotein and V and W proteins have thus evolved by acquiring an additional extension within their common N-terminal region that recruits STAT2. The first 405 N-terminal amino acids of N-terminal region are common to the NiV phosphoprotein and V and W proteins (Fig. 1 B), and these three proteins block interferon signaling by preventing the phosphorylation of STAT1 and by sequestering both STAT1 and STAT2 proteins in the cytoplasm and the nucleus (6,43, 44, 45). The primary binding site of STAT1 was localized between residues 110 and 140 of the N-terminal region by coimmunoprecipitation (6), and this localization was confirmed here by a direct interaction assay. The short section of only 30 amino acids forming this binding site in the NiV phosphoprotein suggests that it is a typical example of a short linear motif that folds upon binding to recruit a protein partner (106) as used by numerous viruses to hijack cellular machineries (107) by mimicking cellular proteins such as STAT1 itself, which uses a short linear motif within its transactivation domain to interact with the TAZ1 domain of the CBP/p300 transcriptional coactivator protein (108).

The Y116A mutant and the H117A/D118A/H119A triple mutant lost their ability to bind STAT1. Similarly, the mutant G121E prevents the accumulation of phosphorylated STAT1 in the nucleus of infected cells. Surprisingly, the G125E mutant, which was unable to inhibit IFN signaling and failed to coprecipitate STAT1 and STAT2 (6,46), shows a similar binding affinity for STAT1 as the WT protein by ITC measurement. By contrast, the G135E mutant conserves the ability to block the accumulation of phosphorylated STAT1 in the nucleus while it also fails to coprecipitate STAT1. Because the affinity between the phosphoprotein and STAT1 is weak, it is possible that the coprecipitation with STAT1 also requires the presence of STAT2, as found for measles virus (105), and that the mutations of Gly125 and Gly135 perturb the cointeraction between STAT1 and STAT2 rather than the direct interaction with STAT1. It is also possible that the difference in affinity, although small, is sufficient to produce an effect in vivo and that interactions with WT and the G125E variant are influenced differently by intracellular conditions such as molecular crowding.

In summary, our results revealed that the NiV phosphoprotein behaves as a nanometer-size octopus-like protein that uses its long flexible arms to expose binding sites for multiple viral and cellular partners and to orchestrate their participation in virus replication, hijacking cellular machines, and interfering with the interferon response.

Accession codes

Coordinates and structure factors have been deposited in the Protein Data Bank (PDB) under accession code PDB: 4GJW. SAXS data have been deposited in the Small Angle Scattering Biological Data Bank under accession code SASDHT4.

Author Contributions

M.R.J., F.Y., J.-M.B., V. Volchkov, M.B., and M.J. conceived the experiments. M.R.J., F.Y., G.C., E.C., C.M., V. Volchkova, and J.-M.B. performed the experiments. M.R.J., F.Y., C.M., N.T., J.-M.B., V. Volchkov, M.B., and M.J. analyzed and interpreted the data. M.R.J., J.-M.B., M.B., and M.J. wrote the manuscript.

Editor: Jill Trewhella.