Ebolavirus VP35 Coats the Backbone of Double-Stranded RNA for Interferon Antagonism

Abstract

Recognition of viral double-stranded RNA (dsRNA) activates interferon production and immune signaling in host cells. Crystal structures of ebolavirus VP35 show that it caps dsRNA ends to prevent sensing by pattern recognition receptors such as RIG-I. In contrast, structures of marburgvirus VP35 show that it primarily coats the dsRNA backbone. Here, we demonstrate that ebolavirus VP35 also coats the dsRNA backbone in solution, although binding to the dsRNA ends probably constitutes the initial binding event.

TEXT

Upon the recognition of double-stranded RNA (dsRNA), host pattern recognition receptors (PRRs) such as RIG-I, MDA-5, LGP2, and TLR-3 activate a signaling cascade that results in the generation of type I interferons (IFN) in infected and neighboring cells. In response, certain viruses, including the filoviruses, have evolved to encode specific proteins that counter innate immune detection by binding to and shielding their dsRNA from the PRRs (1–7). Indeed, the survivability of a filovirus infection is linked to the ability of the host to respond to pathogen-associated molecular patterns like dsRNA in order to mount a strong immune response in the first few days after exposure (8–10).

The two major genera within the family Filoviridae are Ebolavirus and Marburgvirus. The genus Ebolavirus contains five viral species (Ebola virus [EBOV], Reston virus [RESTV], Sudan virus, Bundibugyo virus, and Taï Forest virus). The genus Marburgvirus contains only one viral species, termed Marburg virus (MARV) and its variant Ravn (11).

In all filoviruses, the viral protein that inhibits the IFN response by binding dsRNA is termed VP35 (12), and this function is encoded in its C-terminal ∼130-amino-acid dsRNA-binding domain (RBD). The VP35 protein otherwise is multifunctional and plays essential roles in viral transcription, replication, and nucleocapsid assembly (13–16).

Crystal structures of the VP35 RBDs from EBOV, RESTV, and MARV have been determined alone and in complex with dsRNA. These structures have illustrated that the RBD contains two subdomains, of which one is α-helical and the other is primarily β-sheet (1, 4, 5, 17–19). The β-sheet subdomain contains a conserved basic patch that includes K271, R294, K298, R301, R311, and K328 (RESTV numbering). In structures of the complexes, we observe that VP35s from ebolaviruses bind the dsRNA ends, while VP35s from marburgvirus binds along dsRNA, primarily in the middle.

In VP35 structures from ebolaviruses (EBOV and RESTV), two copies of the VP35 RBDs form an asymmetric dimer to bind to each terminus of the dsRNA. One VP35 RBD, termed the “end-capping” molecule, stacks against the terminal bases of the dsRNA, while the paired VP35 RBD, termed the “backbone-binding” molecule, dimerizes with the end-capping molecule and contacts the neighboring sugar-phosphate backbone (4). In the EBOV structure, an 8-bp dsRNA was crystallized. The four bound VP35 RBDs cover the 8 bp completely (5). In the RESTV structure, an 18-bp dsRNA was crystallized. The same two VP35s bind at each end, but the longer expanse of dsRNA backbone in the center remains exposed; no VP35s are bound to it (4).

In contrast, two crystal structures of MARV VP35 RBD-dsRNA complexes demonstrate that the central dsRNA backbone is bound while the ends remain free (1, 20). Masking of the dsRNA backbone may be essential for immune suppression, as it has been demonstrated that PRRs, like RIG-I and MDA5, oligomerize and recognize the dsRNA backbone in a length-dependent manner (21–23). This apparent requirement for complete backbone coating, coupled with absence of coating in the RESTV structure, made us wonder if the ebolaviruses also coat the dsRNA backbone in solution or if ebolaviruses, indeed, only cap the dsRNA ends (4, 5).

To determine if ebolavirus VP35s also coat the backbone of dsRNA, we performed isothermal titration calorimetry (ITC) analysis of RESTV VP35 RBD binding to various dsRNA molecules. The VP35 proteins were expressed and purified in Escherichia coli, and RNA was obtained (IDT Technologies) and duplexed as previously described (1). VP35 RBD and dsRNA were dialyzed overnight against 4 liters of a mixture containing 10 mM Tris (pH 8.0), 200 mM NaCl, and 2 mM tris(2-carboxyethyl)phosphine (TCEP) buffer. ITC experiments were carried out in duplicate with a MicroCal iTC200 instrument (GE). In these experiments, RESTV VP35, at concentrations ranging from 1 to 2 mM, was loaded into the syringe, while dsRNA, at concentrations ranging from 15 to 40 μM, was loaded into the cell. Each experiment consisted of either 16 injections of 2.5 μl each or 32 injections of 1.25 μl each at intervals of 180 s at a reference power of 5 μcal. In each experiment, these injections were preceded by an injection of 0.5 μl, which was removed from the data analysis (standard procedure, as the dispensed volume of the first injection is often inaccurate). The integrated titration peaks were analyzed and fitted with Origin 7.0 software by using either a “one-set-of-sites” or a “two-sets-of-sites” binding model, as appropriate.

RESTV VP35 RBD binds blunt-ended dsRNA with a complex thermodynamic profile involving two separate events, A and B (Fig. 1). For RESTV, addition of a 3′ overhang to the dsRNA results in significant loss-of-binding event A (Fig. 1 and Table 1) and likely represents end capping. Indeed, crystal structures reveal that the 3′ end of the dsRNA butts into the dimer interface of VP35 from both RESTV and EBOV and that the addition of an overhang at the 3′ end would sterically prevent end capping. Binding event B likely represents backbone binding, as it is unaffected by a 3′ overhang and because more copies of VP35 contribute to binding event B when the dsRNA is longer. RESTV VP35 thus appears to be able to both end cap and backbone coat in solution, although in previous crystal structures, only end capping was observed with a stoichiometry of 4:1 (protein-to-dsRNA ratio) in the complex.

Fig 1.

Isotherms for binding of RESTV VP35 RBD molecules to 18-bp dsRNA with blunt ends, 18-bp dsRNA with a 3′ overhang, and 12-bp dsRNA with blunt ends. The top of each panel shows enthalpic heat released during titration, and the bottom panel shows integrated binding isotherms with the proposed fitted model of binding. Binding events A and B are labeled in the leftmost isotherm.

Table 1.

Thermodynamic parameters of binding determined by ITC for binding of RESTV VP35 to dsRNA molecules that vary in length and the presence of a 3′ overhang at each end^a

RESTV VP35 dsRNA	ΔGAb (kcal/mol)	nA	KDA (μM)	ΔGB (kcal/mol)	nB	KDB (μM)	nAB
18 bp	−8.9	0.9	0.3	−7.6	6.9	2.8	7–8
18 bp + 3′ overhang	Unmeasurablec	0.2	N/A	−7.8	7.8	2.0	7–8
12 bp	−9.6	1.2	0.1	−7.5	3.6	3.0	4–5

^aAll values represent averages of two independent measurements. Superscripts A and B refer to events A and B wherever applicable. n represents the number of molecules binding in that event.

^bThe change in Gibbs free energy (ΔG) was determined as follows: ΔG_binding = RTlnK_D (29).

^cThere are too few measured points to confidently determine the ΔG_binding or the K_D of the first binding event (stoichiometry of <0.2).

We crystallized RESTV VP35 RBD (residues 205 to 329) with a 12-bp palindromic dsRNA at a protein-to-dsRNA ratio of 1.2:1. RESTV VP35 RBD-dsRNA complexes crystallize over a period of 2 to 3 days in a mixture of 0.2 M Li₂SO₄, 0.1 M phosphate-citrate (pH 4.2), and 20% (wt/vol) polyethylene glycol 1000. Data were collected at beam line 5.0.3 of the Advanced Light Source (Berkeley, CA) and integrated with the program D*TREK, and the data collection statistics are summarized in Table 2 (24). The structure of the complex was determined by molecular replacement with MOLREP with the RESTV VP35 RBD (PDB 3KS4) as the starting model (25, 26). Model building was performed with COOT, and structure refinement was performed with PHENIX (27, 28). The final refinement statistics are given in Table 2.

Table 2.

Data collection and refinement statistics

Parameter	Valuea
Wavelength (Å)	0.9765
Space group	P212121
a (Å)	55.09
b (Å)	72.87
c (Å)	175.55
Resolution (Å)	39.3–2.7 (2.84–2.70)
α = β = γ (°)	90.0
Total/unique reflections	87,616/20,077
Redundancy	4.4 (4.5)
% Completion	99.7 (100.0)
I/σ	9.0 (1.9)
Rsym (%)b	9.0 (57.4)
Rcrystc/Rfreed	0.209/0.262
Root mean square deviations
Bonds (Å)	0.007
Angles	1.06
Dihedrals	15.09
Average B values (Å2)
Main chain	41.5
Side chains and water	48.9
All atoms	45.8
Ramachandran plot
Most favored region (%)	94.6
Additional favored region (%)	5.4

^aValues in parentheses are for the highest-resolution shell.

^bR_sym = ΣΣ_i| I_i − <I> |/Σ<I>, where <I> is the mean intensity of the n reflections with intensities I_i and common indices h, k, and l.

^cR factor = Σ_hkl||F_obs|−k|F_cal|/Σ_hkl|F_obs|, where F_obs and F_cal are observed and calculated structure factors, respectively.

^dFor R_free, the sum is extended over a subset of reflections (∼5.0%) excluded from all stages of refinement. Of a total of 20,077 reflections, 1,016 were used for R_free calculations.

The diffraction data were indexed in orthorhombic space group P2₁2₁2₁ with two protein-RNA complexes in the asymmetric unit, each of which contains two RBDs bound to one molecule of dsRNA. Residues 208 to 329 in each copy of the RBD and base pairs 1 to 12 of one dsRNA and 1 to 9 of the other are visible in electron density maps. The N-terminal three residues of each RBD (205 to 207) and base pairs 10 to 12 of the second dsRNA are disordered.

Interestingly, this crystal structure shows that the two RBDs together form one end-capping dimer while the other end of the dsRNA and the central backbone remain unbound (Fig. 2A). This crystal structure suggests that the initial event in dsRNA recognition by ebolavirus VP35 may be the formation of the dimeric end-capping/backbone binding pair on one end rather than, for example, the binding of one end-capping molecule to each end before the attachment of any backbone-binding copies adjacent to them. The previous RESTV complex may thus represent the second event in dsRNA recognition, in which a second end-capping/backbone binding pair caps the other end (4) (Fig. 2B).

Fig 2.

dsRNA recognition events observed in different filovirus VP35-dsRNA complexes. (A) Crystal structure of the RESTV VP35 RBD in complex with a 12-bp dsRNA. The end-capping monomer (number 1 in panel C) is shown in yellow, and the backbone-binding member of the dimer (number 2 in panel C) is shown in teal. (B) Previous crystal structure of the RESTV VP35 RBD in complex with an 18-bp dsRNA (PDB accession no. 3KS8). (C) Continuous-coating crystal structure of the MARV VP35 RBD determined in complex with a 12-bp dsRNA (PDB accession no. 4GHA). Only backbone binding interactions are observed in the crystal structure (numbered 2 to 5, different shades of blue and green). An end-capping interaction (yellow), modeled from RESTV structures by the superimposition of the dsRNA and backbone-binding copy number 2 indicates that no steric clash would occur if MARV also bound an RBD on the dsRNA end (1).

ITC performed on RESTV VP35 suggests that in solution, both end capping and backbone coating occur when sufficient protein copies are available (ITC events A and B, respectively) and that multiple copies of the VP35 RBD can exhibit the backbone-binding function. These results, together with the crystal structures, suggest that for RESTV, and perhaps for ebolaviruses in general, continued backbone coating occurs only after the capping of each available end. Continued coating by ebolavirus VP35s probably resembles the arrangement observed in the MARV structures. Indeed, modeling of the RESTV or EBOV end-capping dimers onto the MARV VP35 continuous-coating structure reveals no steric clashing of the end-capping pair with any continued-coating RBDs (Fig. 2C). The ITC-suggested ability of ebolavirus VP35s to also mask the backbone and the absence of likely structural clashes between the crystallized end-capping VP35s and potential backbone-coating VP35s suggest that ebolavirus VP35, like marburgvirus VP35, is also capable of preventing length-dependent dsRNA recognition by PRRs like RIG-I and MDA-5. The crystal structures, however, appear to reflect the type of interaction that is most stable or most preferred. Higher protein-to-dsRNA ratios, designed to illustrate complete coating and capping, have not yet yielded crystals.

VP35 from the ebolaviruses, but not marburgvirus, thus appears to prefer to end cap as an initial or preferred binding event. Initial end capping may afford ebolavirus VP35 stronger immune suppression than marburgvirus VP35. Indeed, the ebolavirus VP35s have an overall higher affinity for dsRNA and more efficient antagonism of the IFN response than does marburgvirus VP35 (1).

Protein structure accession number.

Coordinates and structure factors for the complex of RESTV VP35 with the 12-bp palindromic dsRNA have been deposited in the Protein Data Bank (PDB) under accession number 4LG2.