The RIG-I ATPase domain structure reveals insights into ATP-dependent antiviral signalling

Abstract

RIG-I detects cytosolic viral dsRNA with 5' triphosphates (5'-ppp-dsRNA), thereby initiating an antiviral innate immune response. Here we report the crystal structure of superfamily 2 (SF2) ATPase domain of RIG-I in complex with a nucleotide analogue. RIG-I SF2 comprises two RecA-like domains 1A and 2A and a helical insertion domain 2B, which together form a 'C'-shaped structure. Domains 1A and 2A are maintained in a 'signal-off' state with an inactive ATP hydrolysis site by an intriguing helical arm. By mutational analysis, we show surface motifs that are critical for dsRNA-stimulated ATPase activity, indicating that dsRNA induces a structural movement that brings domains 1A and 2A/B together to form an active ATPase site. The structure also indicates that the regulatory domain is close to the end of the helical arm, where it is well positioned to recruit 5'-ppp-dsRNA to the SF2 domain. Overall, our results indicate that the activation of RIG-I occurs through an RNA- and ATP-driven structural switch in the SF2 domain.

Figure 1

Structure of the mouse RIG-I SF2 domain. (A) Front view of RIG-I^SF2 along the nucleotide-binding cleft. RIG-I^SF2 consists of three domains: SF2 domain 1A (yellow) and SF2 domain 2A (green) are connected by a short linker helix (grey) and form the conserved ATP-binding and hydrolysis core. The helical bundle domain 2B (red) is a specific feature of RIG-I/FANCM/Hef helicases, indicating that it is involved in double-stranded nucleic acid binding or translocation. An unusual arm (orange), unique to RIG-I-like receptors, reaches from domain 2 across domain 1 and stabilizes the observed ‘open’ conformation. (B) Top view of RIG-I, coloured as in (A). (C) Close-up view of the helical arm and its elbow (orange), embracing the helical protrusion from domain 1A, with a hydrophobic interface. (D) Structure-based sequence alignment of selected RIG-I-like receptors with highlighted conserved residues and annotated motifs. The secondary structural elements are shown on top of the alignment. AMP-PNP, adenosine 5′-(β,γ-imido)triphosphate; RIG-I, retinoic acid inducible gene I; SF2, superfamily 2.

Figure 2

Human RIG-I^SF2 solution structure. (A) Experimental X-ray scattering intensities of human RIG-I^SF2 (blue dots) and calculated scattering intensities from the crystal structure of mouse RIG-I^SF2 (black line), indicating that the mouse RIG-I^SF2 crystal structure is a good model for the human RIG-I^SF2 solution structure. (B) Averaged small-angle X-ray scattering envelope obtained from 16 independent ab initio models from human RIG-I^SF2 (grey) shows the three-lobed ‘C’ shape, with the model of the crystal structure of mouse RIG-I^SF2. Figures and docking were generated using the Situs package (Wriggers, 2010) and University of California, San Francisco Chimera (Pettersen et al, 2004). RIG-I, retinoic acid inducible gene I.

Figure 3

Conserved motifs and mutational analysis. (A) Stereo view of the structure of RIG-I^SF2 with conserved functional motifs shown. (B) Close-up view of the adenosine 5′-(β,γ-imido)triphosphate (AMP-PNP; colour-coded sticks: green, carbon; red, oxygen; blue, nitrogen; orange, phosphorus)-binding moiety with highlighted hydrogen bonds of the adenine recognition site. The protein is shown as colour-coded sticks with yellow carbons. (C) Close-up view of the motifs mutated in this study. Mutated side chains are annotated and shown as sticks. (D) ATP hydrolysis activity of RIG-I SF2 domain mutants. Mixed E374Q+R731A represents a 1:1 mixture of single mutants E374Q and R731A. Plotted bars: mean±s.d. (n=3–6). ND, not detectable; RIG-I, retinoic acid inducible gene I; SF2, superfamily 2; WT, wild type.

Figure 4

Formation of an ATP hydrolysis site requires a structural switch. (A) Comparison of RIG-I^SF2 (open conformation) with DEAD-box enzyme VASA (closed, RNA-bound conformation). Domains 1A are superimposed, showing the strongly differing orientation of domain 2 in RIG-I, stabilized by the helical arm. (B) Model of the RIG-I^SF2 closed conformation, generated by superimposing domain 2A on the structure of VASA domain 2A. All helicase motifs align properly, and this conformation positions QQ and R motifs opposite motifs Ic and IIa, where they are ideally placed to grip dsRNA (brown). (C) Electrophoretic mobility shift analysis of full-length (FL) RIG-I, RD and SF2 domain (SF2D) with 5′-ppp-dsRNA. The RNA-binding affinities of RD and FL are comparable and higher than that of SF2 domain, showing that RD provides the main binding affinity of RIG-I. (D) ATP hydrolysis activity of RIG-I (200 nM) activated with 5′-ppp-dsRNA (25mer), dsRNA (25mer) or 5′-ppp-RNA–DNA hybrid. Plotted bars: mean±s.d. (n=7–10). dsRNA, double-stranded RNA; 5′-ppp, 5′ triphosphates; RD, repressor domain; RIG-I, retinoic acid inducible gene I; SF2, superfamily 2.

Figure 5

Proposed model for RIG-I activation by a conformational switch in the SF2 domain. RD binds to dsRNA with 5′ triphopshates (5′-ppp; magenta spheres) and might recruit it to the SF2 domain. RNA and ATP binding switches SF2 into signal-on conformation by gripping RNA between motifs Ic/IIa on domain IA and R/QQ on domain IIB. The position of the helical arm with the short linker to RD might allow RD to bind 5′-ppp-RNA ends cooperatively with SF2. The precise position of RD, which from our structure might bind on either side of the RNA duplex (both possibilities are shown), and the position of activation and recruitment domains (CARDs) in signal-off and -on states, remain to be determined. dsRNA, double-stranded RNA; RD, repressor domain; RIG-I, retinoic acid inducible gene I; SF2, superfamily 2.