Crystal Structure of Marburg Virus VP40 Reveals a Broad, Basic Patch for Matrix Assembly and a Requirement of the N-Terminal Domain for Immunosuppression

ABSTRACT

Marburg virus (MARV), a member of the filovirus family, causes severe hemorrhagic fever with up to 90% lethality. MARV matrix protein VP40 is essential for assembly and release of newly copied viruses and also suppresses immune signaling in the infected cell. Here we report the crystal structure of MARV VP40. We found that MARV VP40 forms a dimer in solution, mediated by N-terminal domains, and that formation of this dimer is essential for budding of virus-like particles. We also found the N-terminal domain to be necessary and sufficient for immune antagonism. The C-terminal domains of MARV VP40 are dispensable for immunosuppression but are required for virus assembly. The C-terminal domains are only 16% identical to those of Ebola virus, differ in structure from those of Ebola virus, and form a distinct broad and flat cationic surface that likely interacts with the cell membrane during virus assembly.

IMPORTANCE Marburg virus, a cousin of Ebola virus, causes severe hemorrhagic fever, with up to 90% lethality seen in recent outbreaks. Molecular structures and visual images of the proteins of Marburg virus are essential for the development of antiviral drugs. One key protein in the Marburg virus life cycle is VP40, which both assembles the virus and suppresses the immune system. Here we provide the molecular structure of Marburg virus VP40, illustrate differences from VP40 of Ebola virus, and reveal surfaces by which Marburg VP40 assembles progeny and suppresses immune function.

INTRODUCTION

Marburg virus (MARV) and the related Ebola virus (EBOV) both belong to the Filoviridae family. These are enveloped viruses with a nonsegmented single-stranded and negative-sense RNA genome. Genomes of Marburg viruses differ from those of ebolaviruses by more than 50% at the nucleotide level (1). MARV causes severe and rapidly progressing hemorrhagic fever in humans and nonhuman primates, with lethality ranging from 25% to over 90% depending on the geographic location and viral strain (2). The 19-kb MARV genome encodes seven structural proteins. Each of these proteins is of critical importance, and most are known to perform multiple functions during the viral life cycle.

VP40 is the Marburg virus matrix protein, which builds the protein shell underneath the viral envelope and confers the hallmark filamentous morphology to the Marburg virion. VP40 alone is able to induce the assembly and budding of filamentous virus-like particles (VLPs), which resemble authentic virions (3, 4). The intracellular distribution of filovirus VP40 varies during the progression of the viral life cycle (5, 6), as it orchestrates the distribution of the other viral components and viral assembly (6, 7). The majority of it, however, traffics to and associates with the cellular membrane (5, 7, 8). Within the virion, VP40 interacts with both the lipid envelope and the core nucleocapsid complex, which contains the NP, VP35, L, VP30, and VP24 proteins (9). VP40 also interacts with the GP viral surface protein. Indeed, coexpression of VP40 and GP alters the distribution pattern of GP, leading to GP accumulation at the VP40-positive basolateral clusters at which budding occurs (7). MARV VP40 (mVP40) and EBOV VP40 (eVP40) are 34% identical (49% homologous) in amino acid sequence, with differences concentrated in the C-terminal domains (CTDs).

mVP40 has been shown to be immunosuppressive, antagonizing signal transduction from the interferon (IFN) receptor. A dysregulation of the host immune response is characteristic of primate MARV infections (10). In EBOV, VP24 antagonizes signal transduction from the interferon receptor by interacting with karyopherin α (11–14). In MARV, instead of VP24, the VP40 protein antagonizes signal transduction by inhibiting the phosphorylation of JAK1, TYK1, STAT1, and STAT2 under the stimuli of both type I and type II IFNs (15). However, the mechanism by which MARV VP40 exerts this immunosuppressive effect is not yet understood.

Here we describe the crystal structure and biochemical analysis of MARV VP40. We found that MARV VP40 (mVP40) exists as a dimer in solution and, like eVP40, is folded into N- and C-terminal domains. The N-terminal domain of mVP40 is similar in structure to that of eVP40, and the dimer interface it assembles is critical for matrix assembly and budding. The C-terminal domain, however, is more loosely folded than that of eVP40 and features an extended, highly basic patch covering one side. This basic patch is larger and flatter than that of eVP40, and residues within it are essential for matrix assembly and budding. In our analysis of the separate immunosuppressive role of MARV VP40, we found that this function can be attributed to the N-terminal domain (NTD) of mVP40 and is likely a function of the oligomeric ring structure that the NTD assembles.

MATERIALS AND METHODS

Expression of wild-type and mutant MARV VP40.

mVP40 was expressed with an N-terminal 6×His tag in the pET46 Ek/LIC expression vector (Novagen). Site-directed mutagenesis of mVP40 was performed by overlap extension (16), using a plasmid encoding wild-type VP40 as the template.

Rosetta 2 Escherichia coli cells were transformed with the expression vector harboring the gene encoding mVP40. Liter-volume cultures of cells were inoculated with an overnight culture and grown in LB medium (RPI). Cells were grown to mid-log phase (up to an optical density at 600 nm of 0.4) at 37°C, and the culture was cooled at 25°C for 30 min. Protein expression was subsequently induced by addition of 0.3 mM IPTG (isopropyl-β-d-thiogalactopyranoside) (final concentration) with vigorous shaking at 25°C. Following induction for 18 h, cells were harvested by centrifugation.

Expression of selenomethionine-derived MARV VP40.

Hexahistidine-tagged mVP40 contains 10 methionine residues among 317 amino acids. Selenomethionine-labeled mVP40 protein was prepared for phasing by expression of mVP40 in Rosetta 2 (DE3) cells carrying pET46-MARV_VP40. Cells were grown in M9 medium containing 2 mM MgSO₄, 0.1 mM CaCl₂, 0.2% glucose, 1 μg/ml riboflavin, 1 μg/ml thiamine, 1 μg/ml nicotinamide, 1 μg/ml pyridoxine, and 0.04 mg/ml l-amino acids (except Met), plus 100 μg/ml of ampicillin. Cells were grown to mid-log phase at 37°C and cooled at 25°C for 30 min. At mid-log phase, immediately prior to induction, the medium was supplemented with amino acids as follows: 0.1 mg/ml l-Thr, l-Lys, and l-Phe; 0.05 mg/ml l-Leu, l-Ile, and l-Val; and 0.04 mg/ml seleno-l(+)-methionine. Protein expression was induced by adding 0.3 mM IPTG, and growth of the culture was continued overnight. Cells were harvested and frozen at −20°C.

Purification of MARV VP40 protein.

Cells were thawed and suspended in lysis buffer (20 mM Tris-HCl, 500 mM NaCl, 20 mM imidazole, pH 8.0) supplemented with a protease inhibitor cocktail (Roche) and lysed by microfluidization (Microfluidics). The cell extract was centrifuged at 15,000 × g for 1 h, and the supernatant was incubated with Ni²⁺ resin (Qiagen), preequilibrated with lysis buffer, at 4°C for 30 min. The Ni²⁺ resin was washed with more than 15 column volumes of lysis buffer, and the 6×His-tagged mVP40 was eluted with elution buffer (20 mM Tris-HCl, 500 mM NaCl, 300 mM imidazole, pH 8.0). CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate} (10 mM) was added to the protein solution after elution, the protein was loaded onto a size exclusion column equilibrated with buffer (20 mM Tris-HCl, 500 mM NaCl, 10 mM CHAPS, pH 8.0), and fractions containing mVP40 were collected. After size exclusion chromatography was performed, mVP40 was dialyzed against an ion exchange buffer (10 mM Tris-HCl, 200 mM NaCl, 10 mM CHAPS, pH 8.0) at 4°C for 3 h and loaded onto a Mono S ion exchange column (GE Healthcare). mVP40 was eluted using a 200 mM to 500 mM linear gradient of NaCl. The eluted protein solution was dialyzed against size exclusion buffer, concentrated by centrifugal filtration (Amicon) to 3 mg/ml, and used for crystallization. mVP40 samples used for biophysical experiments were equilibrated to 20 mM MES buffer at pH 6.0, after elution from Ni resin.

Size exclusion chromatography with inline multiangle laser light scattering (SEC-MALS).

Purified proteins (typically 0.2 to 0.4 mg total) were loaded onto a Superdex 75 10/300 gel filtration column (GE Healthcare) preequilibrated with buffer (20 mM HEPES-NaOH [pH 8.0], supplemented with 300 mM NaCl). Following elution from the column, the fast protein liquid chromatography (FPLC) system was connected in-line with a miniDAWN TREOS followed by an Optilab T-rEX refractometer (Wyatt Technologies). Data processing and absolute molecular mass calculations were performed using ASTRA software (Wyatt Technologies).

Crystallization of MARV VP40.

Initial screening of crystallization was carried out using an Oryx8 crystallization robot (Douglas Instruments) and commercially available crystal screening kits on MRC vapor diffusion crystallization plates. Minuscule crystals were observed under diverse conditions, including the presence of polyethylene glycol (PEG), salt, alcohol, and other organic precipitants. Microcrystals grown under ethanol conditions were optimized, and crystals that were suitable for diffraction studies were ultimately obtained in 17.5% ethanol–75 mM NaCl–10 mM CHAPS–100 mM Tris-HCl (pH 7.7) at 296 K by hanging drop vapor diffusion.

Diffraction and data collection.

Once crystals grew to equilibrium, the crystal tray was moved to 4°C so that crystals could be harvested at 4°C to avoid damage caused by the evaporation of alcohol. After harvesting, crystals were soaked in a solution containing optimized reservoir solution supplemented with 20% MPD for a few seconds and transferred to liquid nitrogen for freezing.

Multiwavelength anomalous dispersion (MAD) data were collected to 3.11-Å resolution at beamline 23-ID-B of the Advanced Photon Source (APS) (Table 1), Argonne National Laboratory. Crystals were mounted in a cryo-stream of nitrogen at 100 K, and diffraction data were collected through the long axis of rod-shaped crystals using a 20-μm-by-20-μm-diameter mini-beam. Diffraction data were recorded on a Mar Mosaic 300 charge-coupled-device (CCD) detector (Rayonix). The peak wavelength of a Se-Met crystal was based on an X-ray absorption fine structure (XAFS) scan. Complete data sets at the peak (0.9794 Å) and reflection edge (0.9795 Å) were collected from one crystal. Single-wavelength anomalous dispersion (SAD) data were collected to 2.81 Å from another crystal using the same protocol (Table 1). The diffraction data were indexed, integrated, and scaled using HKL2000 (17).

TABLE 1.

Crystallographic data collection and refinement statistics

Parameter	Value(s) for indicated mVP40 anomalous dispersion assaya
	Se-MAD		Se-SAD
	Peak	Inflection
Data collection
Wavelength (Å)	0.9794	0.9795	0.9794
Resolution limits (Å)	50.0–3.11 (3.16–3.11)	50.0–3.11 (3.16–3.11)	50.0–2.81 (2.96–2.81)
Rsymb	0.15 (0.531)	0.145 (0.398)	0.149 (0.530)
Rpimc	0.110 (0.179)	0.115 (0.322)	0.085 (0.313)
I/σI	16.4 (3.9)	13.4 (2.6)	6.9 (2.4)
No. of unique reflections	22,849	21,880	25,034
Completeness	99.9 (99.3)	95.5 (71.1)	99.5 (99.2)
Redundancy	6.3 (6.0)	5.5 (2.9)	4.0 (3.9)
Wilson B (Å2)	24.2	29.8	53.3
Refinement
Resolution (Å)			29.7–2.81
Rwork/Rfreed			0.202/0.260
No. of atoms
Protein			4,040
Water			35
B-factor (Å2)
Mean			23.9
Protein			23.9
Water			23.5
RMS deviationse
Bond length (Å)			0.0112
Bond angle (°)			1.54
Ramachandran plotf
Favored (%)			97.0
Allowed (%)			3.0
Outliers(%)			0.0

^aSe-MAD space group, I222; Se-SAD space group, I222; Se-MAD cell dimensions (Å), a = 96.0, b = 106.8, and c = 128.4; Se-SAD cell dimensions, a = 95.8, b = 107.3, and c = 127.7. Data for the highest-resolution shells are shown in parentheses.

^bR_sym = Σ |I_i − <I_i>|/Σ I_i, where I_i is the observed intensity and <I_i> is the average intensity obtained from multiple observations of symmetry-related reflections.

^cR_pim, precision-indicating and multiplicity-weighted R_merge.

^dR_work = Σ _hkl ||F_obs| − |F_calc||/Σ _hkl |F_obs|. Five percent of the reflections were excluded for the R_free calculation. obs, observed; calc, calculated.

^eRMS, root mean square.

^fData are based on the criteria of MolProbity.

Crystal structure determination.

The structure was solved using the 2W-MAD protocol of Auto-Rickshaw (18), and input diffraction data were prepared using programs of the CCP4 suite (19). Fluorescent antibody (FA) values were calculated using the program SHELXC (20). Based on an initial analysis of the data, the maximum resolution for substructure determination and initial phase calculation was set to 3.6 Å.

All of the 17 heavy atoms requested were found using the program SHELXD (21). The correct hand orientation for the substructure was determined using the programs ABS (22) and SHELXE (20), and the occupancy of all substructure atoms was refined and initial phases were calculated using the program MLPHARE (19). The 2-fold non-crystallographic-symmetry (NCS) operator was found using RESOLVE (23). Density modification, phase extension, and NCS averaging were performed using the program DM (42). A partial alpha-helical model was produced using the program HELICAP (43), and 525 of 622 residues in the asymmetric unit were built using BUCCANEER (24), starting from the HELICAP model. Model refinement was performed using Refmac5 at 2.8-Å resolution and SAD data. Manual model correction was performed using COOT (25), and refinement procedures were iterated until convergence.

VLP budding assay.

Budding of virus-like particles (VLPs) into cell supernatants was detected by Western blot analyses. Wild-type mVP40 and mutant mVP40 bearing a Strep-Tag were cloned into pTriEx-5 (Novagen) and transfected into cells using polyethyleneimine (PEI). VLPs were harvested 48 h posttransfection. Cell culture medium was spun down at 3,000 × g for 5 min to pellet any cells out of the media. Supernatants were layered over 20% sucrose–20 mM HEPES (pH 7.4). VLPs were pelleted at approximately 150,000 × g for 1 h at 277 K and then resuspended in phosphate-buffered saline (PBS) buffer supplemented with 1% NP-40, 1% Triton X-100, and 1 mM EDTA. Cell lysates were collected by washing cells twice with PBS followed by lysis in CytoBuster (EMD Biosciences). VLPs and cell lysates were then run on SDS denaturing gels, transferred onto polyvinylidene difluoride (PVDF) Immobilon transfer membranes (Millipore), and probed with an anti-Strep-Tag antibody (AbD Serotec).

Multilamellar vesicle (MLV) sedimentation assay.

POPC [1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphocholine], POPE [1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphoethanolamine], and POPS [1-hexadecanoyl-2–(9Z-octadecenoyl)-sn-glycero-3-phospho-l-serine] were mixed in the molar ratios noted below and dried under N₂ gas for 10 min. In all sedimentation assays, 30% POPE and 30% POPS were used to maintain consistency (unless otherwise stated) and POPC levels were scaled accordingly. Lipids were hydrated with resuspension buffer (25 mM HEPES [pH 7.4] containing 150 mM KCl) and incubated at 37°C for 15 min to facilitate hydration of lipids to form MLVs (26–30). These lipid solutions were then incubated at room temperature with 5 μM mVP40 proteins for 20 min. The total assay volume was 100 μl, and the final lipid concentration was 1 mM. Samples were centrifuged at 48,000 × g for 20 min. During centrifugation, mVP40 bound to MLV sediments, while unbound proteins remained in the supernatant. Supernatant was carefully removed from the pellet, and the pellet was resuspended in an amount of buffer equal to the volume that was removed as the supernatant. Volumes (40 μl) of each supernatant and pellet sample were separated by SDS-PAGE. The protein band was resolved with Coomassie brilliant blue staining, and gel band intensities were analyzed using Image J software. For the quantification, the gel images are first converted to 8-bit images using Image J software. Using the rectangular selection tool, a rectangular region around the gel bands is selected and marked as the first lane. By selecting plot lanes, a graphical representation of the band intensities as peaks is created by the software, and by using the wand tool, the area under the peak for each gel band is obtained. The fraction of proteins associated with MLVs of each lipid composition was calculated by the following ratio: (area under the peak for pellet gel band)/(area under the peak for supernatant + pellet gel bands). Each experiment was performed at least three times to calculate the mean, the standard deviation (or the standard error of the mean where shown), and the P value.

ISRE reporter gene assay.

293T cells (1 × 10⁵) (ATCC) were transfected with 0.25 μg of interferon-stimulated response element (ISRE) firefly luciferase reporter plasmid, 0.005 μg of Renilla luciferase plasmid, and 1.0 μg of the VP40 pTriEx-5 expression plasmids, encoding N-terminal Strep-Tag II. All transfections were performed using polyethylenimine (31), and cells were treated with 1,000 units of beta IFN (IFN-β; PBL Assay Science) at 40 h posttransfection. At 18 h post-IFN treatment, cells were harvested using reporter lysis buffer and analyzed using a dual-luciferase reporter assay system (Promega), performed according to the manufacturer's instructions. Protein expression was detected by Western blot analysis.

Immunofluorescence (IF) and confocal microscopy.

Hemagglutinin (HA)-fusion WT and mutant (T105R, W83R/N148A) mVP40s were cloned into the pCAGGS plasmid. Intact mVP40 K210E mutant was cloned into pCAGGS plasmid. 293 cells were transfected with plasmids expressing wild-type or mutant mVP40. At 24 h posttransfection, cells were fixed with 4% paraformaldehyde, permeabilized with phosphate-buffered saline (PBS) containing 0.1% Triton X-100 for 5 min at room temperature, and incubated with rabbit anti-VP40 antibody (K210E) or anti-HA antibody (WT, T105R, W83R/N148A). Alexa Fluor 488 goat anti-mouse IgG (Molecular Probes) was used to visualize the primary anti-VP40 and anti-HA antibodies. Cells were observed under an LSM780 microscope (Carl Zeiss).

Protein structure accession number.

Coordinates and structure factors have been deposited into the Protein Data Bank under accession code 5B0V.

RESULTS

Crystal structure of MARV VP40.

mVP40 crystallized in space group I222 with two protomers in the crystallographic asymmetric unit (ASU) and diffracted to a resolution of 2.8 Å (Table 1). The construct crystallized was the complete mVP40, containing residues 1 to 303 (Fig. 1 and 2). Of these, the N terminus (residues 1 to 37 in molecule 1 and 1 to 35 in molecule 2) and a few short loops (residues 71 to 72, 156 to 157, and 264 to 267 in molecule 1 and residues 216 to 218, 259 to 260, 264 to 269, and 302 to 303 in molecule 2) were disordered.

FIG 1.

Sequence alignment. Amino acid sequences of mVP40, Zaire ebolavirus (also known as Ebola virus) VP40, and Sudan ebolavirus (also known as Sudan virus) VP40 were aligned using T-COFFEE (39). The figure was generated using ESPript (40). Secondary-structure elements are indicated as follows: α, α-helix; β, β-strand; η, 3₁₀ helix.

FIG 2.

Crystal structure of Marburg virus VP40. (a) The N-terminal domain-mediated dimer of mVP40 is illustrated. In monomer 1, the N-terminal domain (Nter; residues 38 to 176) is shown in blue and the C-terminal domain (Cter; residues 177 to 303) in orange. In monomer 2, the NTD is green and the CTD yellow. The dimer of MARV VP40 forms a butterfly shape, with the wings twisted relative to each other in and out of the plane of the page. The N termini and some residues in surface loops are not observed in the electron density. Connectivity in missing sections is illustrated by dashed lines. (b) Topology diagram of a single protomer of Marburg virus VP40, colored with the NTD (N) in blue and the CTD (C) in orange. Secondary-structure elements are indicated as follows: α, α-helix; β, β-strand; η, 3₁₀ helix.

mVP40 consists of two distinct domains, an N-terminal domain (NTD; residues 36 to 173) and a C-terminal domain (CTD; residues 183 to 303), which are connected by a linker region (residues 174 to 182) (Fig. 1 and 2b). The mVP40 NTD consists of seven β-strands, two α-helices, and three 3₁₀ helices and adopts a β-sandwich-like structure. The CTD consists of six strands and six helices and adopts a small β-sandwich-like structure as well.

The NTDs of mVP40 and eVP40 are 42% identical in sequence and adopt comparable folds, with an root mean square deviation (RMSD) between the C-alpha atoms of 2.4 Å (Fig. 3a). The CTDs of mVP40 and eVP40, however, are only 16% identical in sequence. The CTDs differ more extensively in structure and align with a 5.6-Å RMSD between C-alpha atoms (Fig. 3b). In particular, residues 208 to 221 of the CTD trace a different pathway in mVP40 than in eVP40 and form an extended, flat, basic patch.

FIG 3.

Comparison of mVP40 and eVP40 in NTD and CTD structures. (a) NTDs (Nter) of mVP40 (blue) and eVP40 (white) were superimposed using least-squares (LSQ) superposition by Coot (25). Overall, the β sandwich structure is conserved, with some differences in β1 and loops that connect the β-strands. (b) Superimposed CTDs (Cter) of mVP40 (orange) and Sudan ebolavirus VP40 (white). Note that the CTDs differ more extensively in structure than the NTDs, particularly the two loops containing residues 209 to 222 (left) and 256 to 271 (right), which trace paths in mVP40 different from those traced by the corresponding residues of eVP40. Residues K210, K211, R215, and K218, which contribute to the basic patch of the mVP40 CTD, are illustrated in the stick model. Residues K264, K265, and R266, which contribute to the basic patch but are on the disordered loop, are indicated in parentheses.

mVP40 forms a dimer, and the dimer interface is essential to matrix assembly.

To date, the only oligomer defined for mVP40 is an oligomeric ring (32). In order to determine if mVP40, like eVP40, also forms a dimer in solution (33), we analyzed purified mVP40 using size exclusion chromatography (SEC) coupled with multiangle light scattering (MALS) (SEC-MALS). In these experiments, wild-type mVP40 was determined to be approximately 63 kDa in solution, while a point mutant, T105R, yielded a 34-kDa, likely monomeric VP40 (see Fig. 5c).

FIG 5.

Cellular localization of wild-type and mutant mVP40s. In each set of three images, the top left panel illustrates nuclear staining with DAPI. Alexa 488-conjugated anti-mouse IgG was used to visualize the primary antibodies. (a) Wild-type (WT) Marburg VP40 localizes to the cell membrane and buds VLPs (indicated with yellow arrows). T105R mVP40 fails to localize to and concentrate at the membrane and fails to assemble and bud VLPs. T105R mVP40 instead stains more diffusely in the cytoplasm. W83R/N148A mVP40 fails to assemble and bud VLPs, does not accumulate at the membrane, and also stains more diffusely in the cytoplasm. (b) Western blot of cell lysates and purified VLPs from transfected cells. The WT strain buds VLPs. Both the T105R and W83R/N148A mutants failed to be released from the cell into the supernatant as VLPs. (c) Multiangle light scattering analysis of the wild type (WT) and mVP40 mutants analyzed by size exclusion chromatography is illustrated. The absolute molecular mass distributions are shown as horizontal lines (WT, 63 kDa; W83R/N148A, 68 kDa; T105R, 36 kDa), with the corresponding light-scattering curves shown as solid lines. WT, W83R/N148A-containing, and T105R-containing mVP40s are shown in black, pink, and blue, respectively. (d) NTD-mediated oligomerization interface model. Two copies of the NTD of mVP40 (pink and purple) are superimposed onto two NTDs of eVP40 in the crystal structure of rearranged, hexameric eVP40 (white). In the lower part of the panel, eVP40 residues W95, E160, R148, and R151, each critical to the hexameric assembly, are shown in white in the stick model. mVP40 residues W83 and N148, which align with eVP40 residues W95 and E160, are shown in magenta. mVP40 residues R136 and R139, which align with eVP40 residues R148 and R151, are shown in purple. The zoomed-in view illustrates the interaction that could be made by mVP40 in this interface.

A potential NTD-to-NTD dimer interface was revealed in the crystal packing of mVP40 that is similar to the dimer interface of eVP40 (Fig. 4a). This interface, however, is smaller than that of eVP40, being approximately 670 Ų. This interface includes strand β1 (residues 40 to 52) and helix α1 (residues 97 to 105) and is assembled mostly by hydrophobic contacts involving residues T40, P41, N42, Y43, L49, D50, Q52, A97, H98, A101, A102, L104, and T105. Three direct hydrogen bonds across the dimeric interface between T40 and N42, between T40 and Y43, and between N42 to N42 of the opposite monomer were noted (Fig. 4b). Residue T105 is central to this interface and forms a hydrophobic contact to Y43. Mutation of threonine at this position to arginine (T105R) resulted in primarily monomeric VP40 in solution, as mentioned earlier (Fig. 5c).

FIG 4.

The dimer interface. (a) Marburg VP40 is colored with NTDs in blue and green and CTDs in orange and yellow. Sudan ebolavirus VP40 is illustrated in white and is aligned on the blue mVP40 NTD. The dimer interface (magnified in panel b) is outlined in red. Structures were superimposed by secondary-structure matching (SSM) using Coot (25). (b) Closeup view of the Marburg virus VP40 dimeric interface, in which one monomer is colored blue and the opposing monomer green. The main chains are illustrated in ribbons, with key residues that contribute to this interface illustrated in stick models. (c) Sequence alignment of the corresponding regions is shown below the model. Residue T105 resides in the middle of the hydrophobic pocket and T40 at the base. SUDV, Sudan ebolavirus.

Immunofluorescence assays (IFA) illustrated that, while wild-type mVP40 successfully assembled and budded VLPs from the cell, T105R mVP40 failed to assemble or bud VLPs. Moreover, T105R mVP40 failed to translocate to the plasma membrane and, instead, remained diffuse throughout the cytoplasm (Fig. 5a and b), as did a mutation at the equivalent position in eVP40, L117R (33). Formation of an NTD-mediated dimer appears critical to assembly of Marburg virus-like particles.

For Ebola virus, dimers are thought to assemble into the viral matrix by a conformational change that exposes a separate surface of the NTD involving W95 and E160 (33, 34). These residues aligned in sequence and structure to W83 and N148 of mVP40 (Fig. 5d). A double W83R/N148A mutant was constructed in order to determine if dimers of mVP40 further assemble via an interface involving a similar surface. SEC-MALS revealed that both wild-type mVP40 and W83R/N148A mVP40 formed dimers in solution as expected (Fig. 5c). However, only WT mVP40 assembled and budded VLPs; the W83R/N148A mutant failed to assemble and to bud VLPs (Fig. 5c).

The CTD bears a basic patch essential for membrane targeting and matrix assembly.

mVP40 interacts with the acidic lipid bilayer surface to maintain the structural integrity of the virus and to organize the matrix structure essential for virus budding (35, 36). A significant basic patch was observed on the surface of the CTD, with the basic patch of each mVP40 monomer exposed along a single side of the dimer (Fig. 6a). This patch was primarily derived from two loops that extended in opposite directions. The first basic loop, spanning residues 208 to 221, contained the basic amino acids K210, K211, R215, and K218 (Fig. 3b and 6a). The corresponding loop of eVP40, spanning residues 210 to 233, was not observed in the eVP40 structure. Its sequence, however, was less basic in charge than that of mVP40 and likely pointed in the opposite direction (Fig. 3b). The second loop in the mVP40 basic patch spanning residues 251 to 271, between β9 and β10, contained basic amino acids K259, K264, K265, and R266 and was similar in sequence composition to that of eVP40. The combined basic patch of mVP40 formed an essentially flat surface, while that of eVP40 involved more extensively projecting, flexible, and disordered loop structures.

FIG 6.

Characterization of the CTD basic patch in mVP40. (a) Top view of the electrostatic surface potential of the mVP40 dimer. (b) Side view. The dimer interface is indicated by a dashed line. The electrostatic surface potential was generated by APBS (41) within CueMol (http://cuemol.sourceforge.jp/en/). Basic residues are labeled. (c) Western blot of cell lysates and purified VLPs from transfected cells. The gel was spliced on the solid line. WT and K210R bud VLPs. VLPs budding of K210S is notably reduced. All mutant proteins were expressed in cells, but those mutants which removed the basic charge failed to be released from the cell into the supernatant as VLPs. (d) IFA demonstrates that WT mVP40 translocates to the cell membrane and buds VLPs (indicated with yellow arrows). In contrast, K210E is impaired in membrane trafficking and does not bud VLPs.

To investigate the role of this basic surface in mVP40, we introduced mutations at three different clusters of basic residues. K210E single, R215A/K218A double, and K264A/K265A/R266A triple mutations were made. Each mutant expressed to a level equivalent to the level seen with wild-type VP40, but each failed to assemble and to bud VLPs (Fig. 6c). Additional substitutions at position 210 confirmed the importance of the basic charge. The acidic K210E mutant failed to bud VLPs, and the neutral/polar K210S mutant was significantly reduced in its ability to bud VLPs, but the basic K210R mutant and the wild-type mVP40 mutant were able to assemble and to bud VLPs from transfected cells (Fig. 6c and d).

A multilamellar vesicle (MLV) sedimentation assay was used to assess the effect of mutations on the CTD basic patch and of the dimer interface on the membrane binding ability of mVP40. mVP40 associates with the plasma membrane by interacting with anionic phospholipids in a nonspecific manner, dependent on the anionic charge density (K. J. Wijesinghe and R. V. Stahelin, unpublished data). We tested K210S, K210E, K210R, R215A/K218A, and K264A/K265A/R266A mutant mVP40 in order to investigate the importance of these residues in mediating electrostatic interactions with the plasma membrane. In this assay, POPC:POPE (70:30) vesicles composed of zwitterionic phospholipids were used as control vesicles. Wild-type mVP40 exhibited only a marginal association with MLVs lacking anionic phospholipids but an 85% association with MLVs containing 30% of the anionic lipid POPS (Fig. 7).

FIG 7.

mVP40 association with lipid vesicles containing neutral phospholipids and anionic phospholipids. (a and b) SDS-PAGE of supernatant (SN) and pellet (P) fractions collected from MLV sedimentation assays on lipid compositions appearing above the images. In all liposomes, 30 mol% of POPE (PE) was used, and, where shown, 30 mol% POPS (PS) was used to maintain consistency. Thus, control vesicles consisted of 70 mol% POPC and 30% POPE and, for test vesicles, 30% POPS was added at the expense of POPC. (See Results for clarification of the distinction between panels a and b.) (c) Quantification of MLV sedimentation assay results. The data represent averages of the results of three sedimentation assays, with error bars representing the standard errors of the means. *, P < 0.05.

mVP40 bearing the K210E mutation exhibited a 35% reduction in its association with anionic phospholipid, phosphatidylserine (POPS)-containing MLVs. mVP40 bearing neutral/polar K210S associated with these MLVs at near wild-type levels, and mVP40 bearing basic K210R achieved a wild-type-level association with MLVs (Fig. 7a and c).

mVP40 bearing either the R215A/K218A double mutation at the second site or the K264A/K265A/R266A triple mutation at the third site exhibited a dramatic decrease in association with POPS-containing MLVs (Fig. 7b and c). These results suggest that alanine replacement in the basic patches of the C-terminal domain disrupts critical electrostatic interactions necessary for mVP40 association with POPS-containing membranes (Fig. 7b and c).

T105 is at the center of the dimeric interface of the mVP40. A T105R point mutation resulted in mVP40 that failed to dimerize and was instead monomeric in solution. T105R-bearing mVP40 also failed to associate with anionic phospholipid-containing MLVs (Fig. 7b and c), suggesting that dimerization, in addition to the basic patch, is also important for membrane association. Dimerization may be important for formation of the extended basic surface formed by the pair of VP40 CTDs.

What structure of mVP40 antagonizes interferon?

In addition to its role in matrix assembly and viral budding, MARV VP40 has an additional role in interferon suppression, antagonizing the host innate immune response by blocking the JAK/STAT signaling pathway (15). Valmas et al. (15) recently demonstrated that mVP40 blocks induction of both ISG54 and gamma-activated-sequence (GAS) reporter genes after the stimulation by type I and type II IFN, respectively. We analyzed various truncations of mVP40 in order to determine which domain or structure of mVP40 antagonizes interferon signaling.

Wild-type mVP40 successfully antagonized IFN-β-stimulated enhancement of ISRE reporter gene expression (Fig. 8), as previously reported (15). The CTD of mVP40, when expressed alone, failed to antagonize IFN signaling. In contrast, the NTD of mVP40, when expressed alone, did antagonize IFN signaling, and it did so as well as wild-type, full-length VP40. Thus, we found that the mVP40 NTD is necessary and sufficient for this function (Fig. 8). Interestingly, the NTD, when expressed alone, forms a ring oligomer (32). Hence, the ring oligomer could be linked to IFN antagonism. For eVP40, the ring is assembled by surfaces of the NTD different from the dimer interface and dimerization is not required for ring formation (33, 37). Similarly, we found that T105R-bearing mVP40, which was unable to dimerize in solution, was fully functional in IFN antagonism. Hence, the formation of the dimer, although critical for matrix assembly, is not critical for the IFN antagonism function of mVP40.

FIG 8.

VP40 inhibits ISRE-induced gene expression. In the ISRE reporter assay, 293T cells were cotransfected with wild-type or mutant VP40s and firefly luciferase driven by an ISRE promoter, treated with IFN-β 40 h posttransfection, and assayed for luciferase activity. Data are shown as the mean fold induction ± standard deviation relative to the untreated negative control, for an individual experiment performed in triplicate. The leftmost control represents empty vector in the absence of IFN-β. Ebola virus VP24 (eVP24) is used as a positive control for interferon antagonism (11).

DISCUSSION

The crystal structure of Marburg virus VP40 presented here illustrates that the N-terminal domain is relatively conserved but that the C-terminal domain varies more extensively in structure among the filoviruses. The structure and accompanying mutagenesis reveal that mVP40 forms an NTD-mediated dimer, which is essential for assembly and budding of virus-like particles. A T105R point mutation at the dimer interface yields monomeric mVP40 and disrupts and assembly and budding of mVP40 VLPs. Previous work demonstrated that a T40A point mutation in mVP40 also disrupts budding (38). This crystal structure reveals that T40 resides on the edge of the dimer interface and makes a hydrogen bond across the interface to the main chain of Y43 of the opposing monomer (Fig. 4b).

For eVP40, dimers are thought to further assemble into the matrix via an “oligomerization interface” containing W95 and E160, which is exposed upon electrostatically driven rearrangement of the NTD and CTD at the membrane (33). Mutagenesis of the equivalent residues in mVP40 (W83 and N148) similarly abolishes the budding of VLPs. Rearrangement of the mVP40 CTD away from the NTD would be necessary to expose this W83- and N148-containing surface of the NTD. Hence, mVP40 probably undergoes structural rearrangement to mediate matrix assembly at the membrane similarly to that of eVP40.

The crystal structure of mVP40 also reveals a single basic patch on the upper surface of the CTDs in the dimer. This basic patch is larger and flatter in mVP40 than in eVP40. Further, the polypeptides that form this surface in MARV individually trace outward in opposite directions, while those of eVP40 point upward and are disordered. As a result, mVP40 forms a broader, flatter basic patch than eVP40. This patch contains three clusters of basic residues, each of which is essential to matrix assembly and budding.

We have further used an MLV sedimentation assay to determine if the basic patch and dimeric interface of mVP40 have an effect on its membrane binding ability. mVP40 associates with membranes containing anionic phospholipids in a highly charge-dependent manner, behaving as a nonspecific anionic charge sensor (Wijesinghe and Stahelin, unpublished). The relatively flat basic patch of dimeric mVP40 provides an ideal surface to mediate electrostatic interactions between the protein and the anionic phospholipids, facilitating recruitment of mVP40 to the plasma membrane. All basic patch mutants in which a cationic charge has been reversed (K210E) or replaced with hydrophobic residues (R215A/K218A and K264A/K265A/R266A) failed to bud VLPs. In membrane association assays, the K210E single-point mutant exhibits a reduction of about 30% in association with POPS-containing MLVs, while double and triple mutants exhibit a more dramatic (∼70%) reduction in association. We suspect that the failure of these mutants to form VLPs occurred because these mutations reduce critical electrostatic interactions necessary to recruit mVP40 to the anionic inner leaflet of the plasma membrane. The mutant behavior in MLV association assays is consistent with the behavior seen in VLP budding assays and strongly supports the idea of the necessity of electrostatic interactions for mVP40 to associate with anionic lipids of the cytoplasmic leaflet of the plasma membrane.

According to immunofluorescence assays (IFA), the monomeric T105R mutant exhibits largely cytoplasmic and nuclear rather than membrane localization. As shown by the MLV assay, T105R-bearing mVP40 cannot robustly associate with anionic phospholipid-containing MLVs. This poor association suggests that mVP40 requires its dimeric structure in order to associate with anionic inner leaflet of the plasma membrane. It is possible that the basic patch of each mVP40 monomer in the dimeric structure has a synergistic effect on membrane association.

Marburg virus VP40 has also been shown to antagonize JAK/STAT signaling, but the molecular mechanism by which this occurs and the structural form of mVP40 responsible for this function were unknown. Here we demonstrate that the NTD of mVP40 is necessary and sufficient for JAK/STAT signaling antagonism. When the NTD of mVP40 is expressed alone, in the absence of the CTD, it assembles into ring-like structures (32). Hence, the ring assembly of mVP40 may be linked to interferon antagonism.

In summary, the 2.8-Å crystal structure of Marburg virus VP40 and additional results of biochemical analysis presented here illustrate the dimeric interface required for virus assembly, assign the domain and, likely, the oligomer of Marburg virus VP40 required for its role in JAK/STAT signal antagonism, and provide a template for biological exploration of the many roles of mVP40 in the Marburg virus life cycle.

Funding Statement

This research was supported by NIH/NIAID grant R44 AI08843 and an Investigators in the Pathogenesis of Infectious Disease award from the Burroughs Wellcome Fund (E.O.S.), a CBBI fellowship supported by NIH T32GM075762 (K.J.W.), NIH grant AI081077 (R.V.S.), Health and Labor Sciences research grants, Japan (Y.K.), and a Grant-in-Aid for Young Scientists from the Japan Society for the Promotion of Science, a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science, and Technology, and PRESTO from Japan Science and Technology Agency (T.N.) and used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under contract DE-AC02-06CH11357.