key: cord-0258753-7oq5fsyz
authors: Yu, Qin; del Valle, Alba Herrero; Singh, Rahul; Modis, Yorgo
title: MDA5 autoimmune disease variant M854K prevents ATP-dependent structural discrimination of viral and cellular RNA
date: 2021-02-01
journal: bioRxiv
DOI: 10.1101/2021.02.01.429133
sha: 6a2697adaee584503c72fbacb2ef4cd7482bbe58
doc_id: 258753
cord_uid: 7oq5fsyz

Our innate immune responses to viral RNA are vital defenses. Long cytosolic double-stranded RNA (dsRNA) is recognized by MDA5. The ATPase activity of MDA5 contributes to its dsRNA binding selectivity. Mutations that reduce RNA selectivity can cause autoimmune disease. Here, we show how the disease-associated MDA5 variant M854K perturbs MDA5-dsRNA recognition. M854K MDA5 constitutively activates interferon signaling in the absence of exogenous RNA. M854K MDA5 lacks ATPase activity and binds more tightly to synthetic Alu:Alu dsRNA. CryoEM structures MDA5-dsRNA filaments at different stages of ATP hydrolysis show that the K854 side-chain forms polar bonds that constrain the conformation of MDA5 subdomains, disrupting key steps in the ATPase cycle-RNA footprint expansion and helical twist modulation. The M854K mutation inhibits ATP-dependent RNA proofreading via a novel allosteric mechanism, allowing MDA5 to form signaling complexes on endogenous RNAs. This work provides new insights on how MDA5 recognizes dsRNA in health and disease.

LGP2 promotes and stabilizes MDA5-RNA complex formation (7, 8) . MDA5 and 34 LGP2 bind dsRNA with a modified DExD/H-box helicase core and a C-terminal domain (CTD) 35 (9-14). The helicase consists of two RecA-like domains, Hel1 and Hel2, and an insert domain, 36 Hel2i, all of which form contacts with phosphate and ribose moieties of both RNA strands. The 37 helicase and CTD, linked by a pair of α-helices referred to as the pincer domain, form a ring 38 around the RNA. Upon binding RNA, RIG-I-RNA oligomers and MDA5-RNA filaments nucleate 39 the assembly of MAVS into microfibrils (15, 16) , which then recruit proteins from the TRAF and 40 TRIM families to activate potent type I interferon and NF-κB inflammatory responses (1, 2, 15) .

its activity in a cell signaling assay. We compared the M854K variant to WT MDA5 and to 91 variants with point mutations at the filament-forming interface engineered to increase or decrease 92 signaling activity ( Fig. 1) (14) . Expression plasmids were transfected into HEK293T cells 93 together with plasmids encoding firefly luciferase under control of the IFN-β promoter and 94 Renilla luciferase under a constitutive promoter. Cells were subsequently transfected with 95 poly(I:C) RNA, a dsRNA analog, to induce MDA5 signaling. IFN-β-dependent gene expression 96 was measured as the ratio of firefly to Renilla luciferase luminescence (14) . The expression level 97 of each MDA5 variant was assessed by Western blotting (Fig. 1) . The luciferase activity of the 98 M854K mutant was 19-fold higher than WT without poly(I:C) stimulation, and 20-40% higher 99 than WT with poly(I:C) stimulation (Fig. 1A) . By comparison, the activity of variant 100 H871A/E875A, shown previously to be hyperactive (14), was 9-fold higher than WT without 101 poly(I:C) stimulation, and 10-30% higher than WT with poly(I:C) stimulation (Fig. 1A) . Deletion 102 of the C-terminal twelve residues (ΔC12), which forms MDA5-dsRNA aggregates instead of 103 helical filaments (14), resulted in no signaling activity without poly(I:C) stimulation and 50% of 104 WT activity with stimulation. The D848K/F849A/R850E mutations, which completely inhibit 105 filament formation but not ATPase activity (14), abolished signaling with and without poly(I:C) 106 stimulation (Fig. 1A) . We conclude that the primary effect of the M854K mutation is to confer 107 constitutive signaling activity in the absence of exogenous RNA. This is consistent with the 108 reported upregulation of type I interferon signaling in a patient harboring the M854K mutation, in 109 the absence of viral infection (28) . The effect of M854K on signaling is distinct from that of gain-110 or loss-of-function mutations at the filament interface, but similar to that of disease mutations in 111 the ATP or RNA binding sites (26). 112 MDA5 recognizes both the structure and length of its dsRNA ligands. The length 113 specificity stems from the RNA binding cooperativity of MDA5, encoded by its filament forming 114 interfaces (4, 5).The sstructural selectivity of MDA5 arises in part from conformational changes 115 coupled to ATP hydrolysis, which promote dissociation from RNAs with imperfect duplexes (14) . 116 A-to-I deamination by ADAR1 weakens base pairing in Alu:Alu mRNA duplexes to prevent 117 autoimmune recognition by MDA5 (19) . The signaling activity of the M854K variant in 118 unstimulated cells was approximately the same as that of WT MDA5 in cells stimulated with 119 poly(I:C) (Fig. 1A) . This suggests that, like previously studied disease-associated MDA5 variants, 120 the M854K variant forms filaments on the relatively short duplexes within cellular RNA (19) . To 121 confirm the importance of the filament interface in activating MDA5 signaling from cellular 122 RNA, we measured the signaling activities of our panel of MDA5 variants in ADAR1 knockout 123 (ADAR1-KO) cells. The mutant with a stabilized filament interface, H871A/E875A, was 124 constitutively active in ADAR1-KO cells, with poly(I:C) stimulation providing no further 125 increase in signaling (Fig. 1B) . The mutant fully deficient in filament formation, stimulation, regardless of MDA5 expression level, and high signaling with poly(I:C) stimulation. 141 Signaling of WT with poly(I:C) increased slightly, but not proportionally with increasing MDA5 142 expression levels (Fig. 1C) . With the M854K variant, however, higher protein expression levels 143 produced higher signaling activity, particularly without poly(I:C) stimulation, when dsRNA was 144 less abundant. The loss-of-function interface mutant, ∆ C12, had no signaling activity stimulation, 145 and proportionally higher signaling with higher protein expression levels with poly(I:C) 146 induction. Doubling the amount of luciferase vector transfected had no significant effect on 147 measured signaling activity, regardless of MDA5 expression level (Fig. S1B) , indicating that the 148 amount of luciferase reporter vector was not limiting in any of the conditions tested. In 149 conclusion, WT MDA5 has an ultrasensitive and cooperative signaling ("On-Off") response that 150 is selective for long dsRNA. M854K and ∆ C12 variants produce more proportional, less 151 cooperative responses, with M854K lacking selectivity and ∆ C12 lacking sensitivity. MDA5 variant M854K lacks ATPase activity and has increased affinity for dsRNA 169 The ATPase activity of MDA5 promotes dissociation from the short (<300-bp) dsRNAs present 170 in the cytosol such as Alu:Alu hybrids. Our cell signaling data imply that, like other gain-of-171 function disease-associated MDA5 mutations (26) for atomic models to be built and refined ( Table 1 ). The highest-resolution structure, with an 226 overall resolution of 2.8 Å and local resolutions up to 2.7 Å, was obtained from ATP-bound 227 filaments (Figs. 3A-C, S2). This structure provided more accurate and precise views of the side 228 chains, ATP, RNA and solvent molecules than previously reported MDA5 structures (Fig. S3) . 229 The ATP-bound M854K MDA5-dsRNA filaments have essentially the same overall structure and 230 helical symmetry as WT MDA5-dsRNA filaments with ATP bound. The Hel1 and Hel2 domains 231 are in the semi-closed state (as defined in (12)) and the helical twist of 75˚ is low (as defined in 232 (14)). With a helical rise of 44 Å, the asymmetric unit spans 14 bp of dsRNA. The M854K variant 233 binds ATP in the same manner as WT MDA5, with a magnesium ion coordinating the βand γ-234 phosphate groups. The most notable difference in the ATP-bound M854K structure is that the 235 mutant Lys854 side chain forms a salt bridge with Glu813 in Hel2, and a hydrogen bond with 236 Ser491 in Hel1, within the same MDA5 protomer (Fig. 3D) . These polar bonds are more 237 constrained in distance and orientation than the weak hydrophobic contacts formed by the Met854 238 side chain in WT MDA5 ( Fig. S4A-B) . Thus, the M854K mutation imposes constraints absent in 239 WT, between the pincer and He1/Hel2 domains. 240 We note that despite incorporating three times more segments in the dominant ADP-AlF 4 -282 bound filament reconstruction than in the equivalent WT reconstruction (14) shown for reference and was reported previously (14) and RNA present rather than an ultrasensitive "On-Off" response. Despite its location in the 391 pincer domain, outside the RNA interface and ATP binding site, the M854K mutation abolishes 392 ATPase activity. Since ATP hydrolysis promotes dissociation of MDA5 from imperfect RNA 393 duplexes, we propose that the M854K variant induces autoimmune signaling due to a lower 394 dissociation rate from endogenous RNAs. The net effect of M854K on signaling is similar to that 395 of disease mutations in the ATP or RNA binding sites (26), or loss of ADAR1 activity (19). 396 Our cryoEM structure of the M854K-dsRNA filament with ATP bound contains structural 397 information up to 2.6 Å resolution, significantly higher than previously published MDA5 filament 398 structures. The structure shows that Lys854 forms polar contacts with the Hel1 and Hel2 domains, 399 imposing constraints on the distance and relative orientation of these two domains. Consistent 400 with this, the filaments had a narrower twist distribution than WT MDA5 filaments with ATP 401 bound. In cryoEM reconstructions of the ADP-AlF 4 transition state, most of the M854K MDA5 402 filament protomers fail to expand their RNA footprint from 14 to 15 base pairs. The twist 403 distribution of ADP-AlF 4 -bound filaments was also broader for M854K MDA5 than for WT. 404 Together, these findings suggest that the M854K mutation prevents MDA5 from maintaining a 405 tight grip of the RNA in the transition state. RNA footprint expansion has been proposed to 406 contribute to discrimination between self-and non-self RNAs, repair of filament discontinuities, 407 and displacement of viral proteins from dsRNA (14, 31, 32) . By partially inhibiting RNA 408 footprint expansion, the M854K mutation will interfere with each these functions. 409 The structure of ADP-bound WT MDA5 filaments completes our structural picture of the 410 MDA5 ATPase cycle (Fig. 6) . In contrast to the ATP-and transition states, helicase motifs Vc The ADP-AlF 4 -bound M854K MDA5 filament dataset was processed in the same way as 556 the ADP-bound dataset except that the optics groups were not split.

The ATP-bound M854K MDA5 dataset was processed in the same way as the ADP-bound 558 dataset except that dose-weighted images were used for CTF estimation. A reference-free two-559 dimensional (2D) class was generated from manual picking as a template for auto-picking. 560 Helical reconstruction was performed using the ATP-bound WT MDA5-dsRNA structure (EMD-561 0024 (14)) as the initial model. 1,068,108 auto-picked segments were subjected to three rounds of 562 2D classification yielding 383,128 particles. Three independent 3D classifications of ATP-bound 563 filament segments did not show any significant differences in twist and helical rise distribution of 564 the selected segments. The ATP-bound segments were 3D-refined, CTF-refined, Bayesian-565 polished and postprocessed following the workflow shown in Fig. S2E Left, Class 167k; Right, Class 62k (see Table 1 ). twist; right, 88˚ twist (see Table 1 ). 

Differential roles of MDA5 and RIG-I helicases 602 in the recognition of RNA viruses

Length-dependent recognition of double-stranded 605 ribonucleic acids by retinoic acid-inducible gene-I and melanoma differentiation-606 associated gene 5

Antiviral immunity via RIG-I-610 mediated recognition of RNA bearing 5'-diphosphates

Cooperative 612 assembly and dynamic disassembly of MDA5 filaments for viral dsRNA recognition

Modis, MDA5 cooperatively forms dimers and ATP-sensitive filaments 615 upon binding double-stranded RNA

MDA5 assembles into a polar helical 617 filament on double-stranded RNA

The innate immune sensor LGP2 619 activates antiviral signaling by regulating MDA5-RNA interaction and filament assembly

Viral RNA recognition by LGP2 and MDA5, and 623 activation of signaling through step-by-step conformational changes

Structural Basis for the Activation of Innate Immune Pattern-Recognition 627

Receptor RIG-I by Viral RNA

Structural basis of RNA recognition and activation by innate immune receptor RIG-I. 630

Structural 632 insights into RNA recognition by RIG-I

Structural Analysis 634 of dsRNA Binding to Anti-viral Pattern Recognition Receptors LGP2 and MDA5

Structural basis for dsRNA recognition, filament formation, and antiviral signal activation 638 by MDA5

Cryo-EM Structures of MDA5-dsRNA Filaments at Different 640 Stages of ATP Hydrolysis

MAVS forms functional 642 prion-like aggregates to activate and propagate antiviral innate immune response

Molecular imprinting as a signal-activation mechanism of the viral RNA 646 sensor RIG-I

MDA5 Governs the Innate Immune Response to SARS-650

CoV-2 in Lung Epithelial Cells

SARS-CoV-2 triggers 653 an MDA-5-dependent interferon response which is unable to control replication in lung 654 epithelial cells

Breaching Self-Tolerance to Alu Duplex RNA Underlies MDA5-Mediated Inflammation

Epigenetic 664 therapy induces transcription of inverted SINEs and ADAR1 dependency

The HUSH complex is a gatekeeper of type I interferon through epigenetic 668 regulation of LINE-1s

Type I interferon-mediated monogenic autoinflammation: The 670 type I interferonopathies, a conceptual overview

RNA editing by ADAR1 leads to 674 context-dependent transcriptome-wide changes in RNA secondary structure

A specific IFIH1 gain-of-function mutation causes Singleton-Merten 679 syndrome

Gain-of-function mutations in IFIH1 cause a 689 spectrum of human disease phenotypes associated with upregulated type I interferon 690 signaling

Genetic and phenotypic spectrum associated with IFIH1 701 gain-of-function

A novel IFIH1 mutation in the pincer domain underlies the clinical 704 features of both Aicardi-Goutieres and Singleton-Merten syndromes in a single patient

Signature in the Italian Aicardi Goutieres Syndrome Cohort: Report of 12 New Cases and 713 Literature Review

New tools for automated high-resolution cryo-EM structure determination in 716 RELION-3

Kinetic mechanism for viral dsRNA length discrimination by MDA5 filaments

ATP-dependent effector-like functions 722 of RIG-I-like receptors

Syndrome: Two Faces of the Same Type I Interferonopathy Spectrum

A time-and cost-efficient system for high-level 727 protein production in mammalian cells

CTFFIND4: Fast and accurate defocus estimation from electron 730 micrographs

Helical reconstruction in RELION

Estimation of high-order aberrations and 734 anisotropic magnification from cryo-EM data sets in RELION-3.1

Beam-induced motion correction for sub-megadalton cryo-EM particles

UCSF Chimera--a visualization system for exploratory research and analysis

Coot: model-building tools for molecular graphics

Macromolecular structure determination using X-rays, neutrons and electrons: 748 recent developments in Phenix

CryoEM data were collected at the MRC-LMB, eBIC at Diamond Light Source (DLS) and the 751

European Molecular Biology Laboratory (EMBL). eBIC is supported by grant EM17434 from the 752

We thank the following facility staff for assistance in 753 cryoEM data collection: Bilal Ahsan, Giuseppe Cannone, Grigory Sharov, Shaoxia Chen and 754 other staff at the MRC-LMB EM Facility

We thank William Schafer (MRC-LMB) and Charles Rice (The Rockefeller University) 756 for providing WT and ADAR1-/-HEK cells

Darling (LMB Scientific Computing) for IT support for image processing. We thank Takanori 758

Emsley and Jude Short for helpful suggestions. We thank Brian Ferguson (Univ

Cambridge) for comments on the manuscript and useful discussions. This work was funded by 760

Wellcome Trust Senior Research Fellowships

Wellcome Trust PhD Fellowship 215378/Z/19/Z to R.S. The work was supported by the NIHR 762

Cambridge BRC (the views expressed are those of the authors and not necessarily those of the 763 NIHR or the Department of Health and Social Care)

collected and analyzed 767 the cryoEM data, and performed the image processing and reconstruction. All authors contributed 768 to model building. Q.Y. refined the atomic models. All authors contributed to the figures. The 769 first draft of the manuscript was