Paramyxovirus V Proteins Disrupt the Fold of the RNA Sensor MDA5 to Inhibit Antiviral Signaling

See allHide authors and affiliations

Science  08 Feb 2013:
Vol. 339, Issue 6120, pp. 690-693
DOI: 10.1126/science.1230949


The retinoic acid–inducible gene I (RIG-I)–like receptor (RLR) melanoma differentiation–associated protein 5 (MDA5) senses cytoplasmic viral RNA and activates antiviral innate immunity. To reveal how paramyxoviruses counteract this response, we determined the crystal structure of the MDA5 adenosine 5′-triphosphate (ATP)–hydrolysis domain in complex with the viral inhibitor V protein. The V protein unfolded the ATP-hydrolysis domain of MDA5 via a β-hairpin motif and recognized a structural motif of MDA5 that is normally buried in the conserved helicase fold. This leads to disruption of the MDA5 ATP-hydrolysis site and prevention of RNA-bound MDA5 filament formation. The structure explains why V proteins inactivate MDA5, but not RIG-I, and mutating only two amino acids in RIG-I induces robust V protein binding. Our results suggest an inhibition mechanism of RLR signalosome formation by unfolding of receptor and inhibitor.

RIG-I–like receptors (RLRs) play a central role in the recognition of viral nucleic acids by the innate immune system (13). RLRs comprise the three family members retinoic acid–inducible gene I (RIG-I), melanoma differentiation–associated protein 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2). RIG-I and MDA5 differentially sense a broad range of viruses (4), and their domain structure consists of N-terminal tandem caspase activation and recruitment domains (CARDs), a superfamily 2 (SF2) adenosine triphosphatase (ATPase), and a C-terminal RNA binding domain (RD). LGP2 lacks CARDs, and its functions remain unclear (5, 6). RIG-I is activated by pathogen-associated nucleic acid patterns such as 5′-triphosphate–containing and double-stranded RNA, whereas MDA5 responds to longer or more complex double-stranded RNA (dsRNA) networks (712). RNA sensing unmasks the CARDs in RIG-I (2, 1315), leading to downstream signaling in a process that involves K63 ubiquitin chains and polymerization of the mitochondrial antiviral-signaling protein MAVS (1618).

Viruses have evolved diverse means to evade innate immune signaling and the antiviral interferon response (19). Paramyxoviruses (e.g., measles, parainfluenza, Sendai and Nipah viruses) synthesize V proteins, which enhance pathogenicity and limit interferon production by targeting MDA5 and LGP2, but not RIG-I (2022). V proteins share the N-terminal domain (NTD) with viral P proteins but contain a distinct, highly conserved C-terminal domain (CTD). The CTD binds the SF2 domain of MDA5 and is necessary and sufficient for inhibiting ATPase activity and cellular aggregation of MDA5 (2023).

To understand a structural mechanism by which a viral inhibitor targets an RLR signaling component, we crystallized and determined the 2.3 Å resolution crystal structure of a complex between parainfluenza virus 5 (PIV5) V protein and the SF2 domain of porcine MDA5 (Sus scrofa ssMDA5) (Fig. 1, fig. S1, and table S1). We obtained crystals of a copurified complex of ssMDA5 SF2 (residues 546 to 808) and PIV5 V protein (residues 168 to 219) only after adding trace amounts of trypsin. This treatment cleaved away several flexible parts due to partial unfolding (see below) that had hampered crystallization despite extensive trials with different V proteins and MDA5 species and constructs.

Fig. 1

The ssMDA5:PIV5 V complex. (A and B) Structure of the ssMDA5:PIV5 V core complex in ribbon representation with annotated secondary structures. The helical insertion domain 2B of ssMDA5 is shown in blue and the superfamily 2 (SF2) ATPase domain 2A in yellow. The CTD of PIV5 V protein is shown in red with Zn2+ ions in gray. PIV5 V protein and MDA5 form a 1:1 complex through interaction of core secondary-structure elements.

Our structure represents the evolutionarily conserved, necessary, and sufficient core of the MDA5:V protein complex (2124). It comprises domains 2A and 2B of the MDA5 SF2 domain and the PIV5 CTD in a 1:1 complex (Fig. 1 and fig. S1). Of particular importance is domain 2A, which together with 1A (not present in our structure) forms the ATP binding site and together with 1A and 2B forms the RNA binding site of RLRs (1315, 25).

MDA5 domain 2A had the "RecA-like fold" typical of ATPases, which harbored a central β sheet with parallel strands in the order β3-β4-β2-β5-β1-β6. To our surprise, although strands β3-β4-β2-β5 formed the expected sheet, β1, β6, and the ATPase motif VI between β5 and β6 were missing, despite being highly conserved in MDA5. Instead, we found the CTD at the expected position of β1 and β6 (Fig. 1 and fig. S1C). The CTD contained a small core with two zinc ions that stabilize a large protruding β hairpin as well as a flanking, protruding loop. The β hairpin of the CTD formed a continuous β sheet with domain 2A in the order β3MDA5-β4MDA5-β2MDA5-β5MDA5-β1CTD-β2CTD, while the loop occupied the usual position of motif VI in SF2 domain–containing enzymes. Thus, our data imply that V protein unfolds the core β sheet of MDA5 and disrupts its ATPase domain by displacing motif VI.

We found three particularly noteworthy contacts between the V protein CTD and the SF2 domain of ssMDA5 (Fig. 2A). First, the β sheet formed between β5MDA5 and β1CTD mediated the central part of the interaction and extended over four residues from MDA5 and four residues from CTD. Second, two contacts at the base and tip of the hairpin provided sequence-specific interactions with MDA5. E174PIV5 formed hydrogen bonds and ion-pair interactions with R803ssMDA5, whereas W179PIV5 at the tip of the β hairpin inserted like a hook into a cavity between domains 2A and B of MDA5 and stacked to G805ssMDA5 at the tip of β5. Additional interactions were formed between the loop of V protein and MDA5 and altogether, the CTD bound to domain 2A via a 1154 Å2 large and highly complementary interface. Notably, R803ssMDA5, G805ssMDA5, W179PIV5, and E174PIV5 were highly conserved among MDA5 and V proteins, respectively (Fig. 2B). Furthermore, LGP2 proteins had arginine and glycine at positions of R803ssMDA5 and G805ssMDA5, whereas RIG-I proteins had leucine instead of arginine and glutamate instead of glycine. Hence, RIG-I proteins cannot form the conserved salt bridge to E174PIV5 and sterically hinder W179PIV5. Thus, our structural results explained why V proteins can bind to MDA5 and LGP2 but not to RIG-I (21, 22, 26).

Fig. 2

Structural details and analysis of the interaction interface. (A) Close-up views of the intermolecular ssMDA52A:PIV5 VCTD and the intramolecular PIV5 VNTD:VCTD interactions (indicated by dashed lines). Domain 2A and the NTD are shown in yellow, the CTD is shown in purple. (B) Structure-based alignments of the interface sequences of selected RIG-I–like receptors (left) and paramyxoviral V proteins (right) with highlighted conserved residues and motifs. Stars indicate residues involved in interface interactions; nonfilled stars indicate residues mutated in this study. Species abbreviations are as follows: Ap, Anas platyrhynchos; Gg, Gallus gallus; Hs, Homo sapiens, Mm, Mus musculus; Ss, sus scrofa. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (C) Western-blot analysis of coimmunoprecipitation experiments (Co-IP) performed with flag-tagged hsMDA5 or hsRIG-I and MV V protein (n = 3). EV, empty vector. Left: Mutations affecting the salt-bridge formation between the conserved glutamate of V protein and arginine of MDA5. Right: Mutation of RIG-I to mimic the critical MDA5 arginine. (D) Comparison of RIG-I2A (PDB code: 3TBK), MDA52A:PIV5 VCTD, and full-length PIV5 V (PDB code: 2HYE).

To test the importance of the observed interactions and validate the structure, we mutated MDA5 and V protein and tested complex formation by coimmunoprecipitations (co-IPs) of human MDA5 (Homo sapiens hsMDA5) and measles virus (MV) V protein coexpressed in human embryonic kidney–293T (HEK-293T) cells (Fig. 2C) and by gel filtration of purified murine MDA5 (Mus musculus mmMDA5) and PIV5 V protein in vitro (fig. S2A). Mutating R806mmMDA5 (equivalent to R803ssMDA5) to leucine (as seen in RIG-I) abolished PIV5 V protein binding in gel filtration and reduced the interaction between hsMDA5 and MV V protein in co-IPs. Likewise, mutating E235MV (equivalent to E174PIV5) to alanine strongly reduced binding to hsMDA5 in co-IPs. A similar substitution with alanine in E174PIV5 also reduced the interaction with mmMDA5 in vitro. These data demonstrated the importance of both sides of the salt bridge and validated the interaction with proteins of four different species. Previous mutational analysis conducted on the basis of sequence conservation is also fully compatible with the crystal structure and highlighted the importance of residues in β1CTD in inactivating MDA5 in living cells (24, 27). Finally, mutating the conserved salt bridge in the MDA5:V protein interface reduced the ability of MV V protein to inhibit activation of the interferon-β promoter via hsMDA5 signaling in living cells (fig. S3).

The sequence and structural similarity to RIG-I (Fig. 2B and fig. S1C) suggested that R803, β1, and β6 of MDA5 domain 2A were mostly buried between motif VI and the helical arm that connects domains 2A and 1A in RIG-I (Fig. 2D). To further validate the importance of R803, we mutated the equivalent residue L714hsRIG-I (or the equivalent L715mmRIG-I) to arginine and analyzed the interaction of the mutated RIG-I with MV or PIV5 V protein (Fig. 2C and fig. S2B). We also analyzed the double mutants L714R,E716GhsRIG-I (or L715R,E717GmmRIG-I) to account for the likely clash between W179PIV5 and the glutamate that RIG-I has instead of the glycine found in MDA5. A single L714hsRIG-I →R mutation induced binding of V protein to hsRIG-I in co-IPs (Fig. 2C). Introduction of a second mutation in mmRIG-I (L715R,E717GmmRIG-I) led to interaction with PIV5 V protein that was stable in gel filtration (fig. S2B), and L714R,E716GhsRIG-I was as efficiently coimmunoprecipitated with MV V protein as MDA5 (Fig. 2C). Thus, we conclude that our model describes the recognition of MDA5 by V proteins sufficiently well that we can engineer robust binding by V proteins into RIG-I by exchanging only two amino acids.

Some V proteins also target host signal transducer and activator of transcription (STAT) proteins to the DDB1-Cul4A ubiquitin ligase complex for proteasomal degradation to disrupt interferon signaling (28). PIV5 V protein bound to DDB1 [Protein Data Bank (PDB) code: 2HYE] is a globular protein with a central β sheet (29) (Fig. 2D). The β hairpin of the CTD in this complex was surrounded by the NTD and did not participate in DDB1 binding (Fig. 2D). Thus, not only was the interaction of V protein with MDA5 fundamentally different from its interaction with DDB1, but V protein also must undergo a large structural change to expose the β hairpin for binding to MDA5 (Fig. 2A). The NTD contains an arginine (R143PIV5) that binds to E174PIV5 and structurally "mimics" R803ssMDA5 (Fig. 2A). We used R143PIV5 to further validate our structural results. If the NTD is only a cocoon before β-hairpin liberation, mutating R143PIV5 should not affect complex formation of V protein with MDA5. Consistently, we found that R143APIV5 formed a highly stable complex with mmMDA5 (fig. S2A). These data support a "spring blade" model for V protein. In the absence of MDA5, the β hairpin is bound by the NTD. In the presence of MDA5, the β hairpin unfolds and disrupts the ATPase domain of MDA5, whereas the NTD might adopt the unfolded conformation seen in P protein (30). In agreement with this, we also observed an increase in the maximum particle size (Dmax) and intramolecular distances of MDA5 in the presence of the PIV5 V protein in small-angle x-ray scattering experiments. This increase is in good agreement with the structural data, because V protein wedged between domains 1 and 2 of MDA5 and likely increased flexibility between domains 1 and 2 and expanded the MDA5 fold (figs. S4 and S5). Taken together, inhibition of MDA5 by V proteins involves unfolding of both the receptor and the inhibitor.

Consistent with the unfolding mechanism, MDA5 and LGP2 bound by coexpressed V proteins lacked or had reduced ATPase activity (21), and titrating purified MV V protein into a solution containing purified hsΔCARD-MDA5, poly(I:C), and ATP progressively inactivated MDA5's ATPase activity (Fig. 3A). In addition, we noticed substantial changes in the way MDA5 bound RNA in the presence of PIV5 V protein (Fig. 3B and fig. S6A). In the absence of V protein, mmMDA5 shifted 24–base pair (bp) RNA predominantly to a single species, with a slight amount of a faster-migrating species at low MDA5 concentrations. V protein strongly increased the amount of faster-migrating species at low MDA5 concentrations. A plausible explanation is that V protein prevents cooperative binding of MDA5 to dsRNA (31, 32) by disrupting the SF2 domain architecture, but still allows noncooperative binding of MDA5 via RD (fig. S6B). Consistently, it has been observed that V protein reduces the size of MDA5-containing oligomers in cells (23).

Fig. 3

Impact of V protein on the ATPase- and the RNA-binding activity of MDA5. (A) The ATPase activity of hsΔCARD-MDA5 (100 nM final) in response to MV V protein. Data represent the mean ATPase activity (±SEM) determined by analyzing the initial linear slopes of the ATPase reactions (n = 3). (B) RNA-binding affinities analyzed by electrophoretic mobility–shift assays of mmMDA5 in the presence and absence of PIV5 V protein (n = 3). The accumulation of a faster-migrating species is marked by a star. (C) mmMDA5 filament formation on poly(I:C), visualized by negative-stain EM in the presence and absence of ATP and PIV5 V protein. (D) Proposed model for inhibition of MDA5 signaling assemblies by paramyxoviral V proteins (29, 30). dsRNA-bound MDA5 filaments might interact with MAVS and induce the formation of MAVS fibrils postulated to be active signal entities (17). V proteins disrupt the MDA5-SF2 architecture by a β-strand–replacement mechanism accompanied by double-unfolding of both V protein and MDA5, resulting in disruption of MDA5:dsRNA filaments and hence of signal transmission of the antiviral response. The structure and interactions of the NTD in the MDA5 complex are only illustrated here and need to be addressed in future studies.

To directly determine whether binding of V protein prevents the cooperative assembly of MDA5 on dsRNA into filaments that are postulated as signaling entities (31, 32), we visualized mmMDA5 and mmMDA5:PIV5 V protein complexes in the presence of poly(I:C) (a synthetic analog of dsRNA) with or without ATP by negative-stain electron microscopy (EM) (Fig. 3C and fig. S7). In the absence of V protein, mmMDA5 robustly formed filaments on poly(I:C), as observed previously (31, 32). Addition of V protein drastically reduced the number of filaments both in the presence and absence of ATP. These data suggest that unfolding of MDA5 by V protein also prevents formation of RNA-bound signaling oligomers (Fig. 3D)—for instance, by disrupting SF2-SF2 or SF2-RD contacts between filament protomers (33).

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Table S1

References (3450)

References and Notes

  1. Acknowledgments: We thank S. Cui and F. Civril for help in the initial stages of the project and K. Lammens, M. Bennett, and N. Fenn for technical advice. We thank O. Berninghausen, C. Ungewickel, and G. Pfeifer for help with EM. We thank the Max Planck Crystallization Facility (Max Planck Institute Martinsried) for screening of crystallization conditions. We thank T. Fröhlich for mass spectrometry analysis, I. Mathes for N-terminal sequencing data, and C. Basquin for support in static light-scattering analysis. We thank the Swiss Light Source (Villingen, Switzerland), the European Synchrotron Radiation Facility (Grenoble, France), and the European Molecular Biology Laboratory (DESY, Hamburg, Germany) for beamtime and on-site support. The data reported in this paper are tabulated in the main paper and in the supplementary materials. Atomic coordinates and structure factors of the reported crystal structure have been deposited in the Protein Data Bank under accession code 4I1S. Details and supplementary materials are available on Science Online. The authors declare no conflict of interest. Correspondence and requests for materials should be addressed to K.-P.H. ( This work is funded by grants from the German Research Council [Deutsche Forschungsgemeinschaft (DFG) GRK1721] to K.-P.H. and G.W., NIH (U19AI083025), the Bavarian government (BioSysNet), and the Excellence Initiative of the German Ministry of Education and Science (Center for Integrated Proteins Science) to K.-P.H. A.K. and K.M.S. were funded by DFG GRK1202.
View Abstract

Stay Connected to Science

Navigate This Article