Characterization of RNA driven structural changes in full length RIG-I leading to its agonism or antagonism

Abstract

RIG-I (retinoic acid inducible gene-I) can sense subtle differences between endogenous and viral RNA in the cytoplasm, triggering an anti-viral immune response through induction of type I interferons (IFN) and other inflammatory mediators. Multiple crystal and cryo-EM structures of RIG-I suggested a mechanism in which the C-terminal domain (CTD) is responsible for the recognition of viral RNA with a 5′-triphoshate modification, while the CARD domains serve as a trigger for downstream signaling, leading to the induction of type I IFN. However, to date contradicting conclusions have been reached around the role of ATP in the mechanism of the CARD domains ejection from RIG-I’s autoinhibited state. Here we present an application of NMR spectroscopy to investigate changes induced by the binding of 5′-triphosphate and 5′-OH dsRNA, both in the presence and absence of nucleotides, to full length RIG-I with all its methionine residues selectively labeled (Met-[ϵ-¹³CH₃]). With this approach we were able to identify residues on the CTD, helicase domain, and CARDs that served as probes to sense RNA-induced conformational changes in those respective regions. Our results were analyzed in the context of either agonistic or antagonistic RNAs, by and large supporting a mechanism proposed by the Pyle Lab in which CARD release is primarily dependent on the RNA binding event.

Graphical Abstract

Graphical Abstract.

INTRODUCTION

The innate immune system provides a first line of defense against pathogens and leads to activation of the adaptive immune system through antigen presentation (1). It is well established that one of the mechanisms through which the innate immune system senses pathogen infection is the detection of foreign nucleic acids, recognizing features conserved among pathogens and commonly called pathogen-associated molecular patterns (PAMPs; i.e. 5′-ppp modification) (2). The pathogen recognition step is initiated by the pattern recognition receptors among which are the cytoplasmic retinoic acid inducible gene-I like receptors (RLRs). These receptors can discriminate between self and viral nucleic acids, triggering an innate immune response (3–5). The RLRs play a key role in effective antiviral defense with two helicases, RIG-I and MDA5 (melanoma associated gene 5) being essential in this process (3,6). Being part of the same receptor family, RIG-I and MDA5 are composed of a modified DExD/H-box ATPase core, a C-terminal domain (CTD), and tandem N-terminal caspase activation and recruitment domains (CARDs) (7–9). Analysis of their respective signaling mechanisms revealed that RIG-I senses primarily short blunt end dsRNA, while the optimal length of dsRNA can be as long as 2000 base pairs (bp) for MDA5 (3,8,10). In both RIG-I and MDA5, the CTD is responsible for initial binding to RNA, and a wealth of crystallographic and recent cryo-EM information for RIG-I has provided a detailed understanding of how 5′-OH, 5′-diphosphate (pp) and 5′-ppp RNA interact with RIG-I’s CTD (11–16). It is clear from these structural studies that the CTD and HEL2 are critically involved in distinguishing host from viral RNA by making different types of contacts to 5′-terminal substituents and adopting different conformations, respectively, while dsRNA binding is sufficient for releasing CARDs for downstream signaling (16). However, despite extensive crystallographic and cryo-EM efforts no structure is available to-date in which the behavior of each CAR domain could be directly visualized in the context of RNA bound FL RIG-I, due to their highly dynamic nature. Moreover, prior to the recent cryo-EM work alternate approaches (HDX-MS, SAXS and FRET) were implemented to better understand the molecular mechanism by which FL RIG-I triggers CARDs release in solution. (17–19). However, while the first two methods suggested that nucleotides (ATP, ADP-AlFx) binding may be required to fully release the CARDs, the recent series of high-resolution cryo-EM structures of RIG-I in complex with host and viral RNA ligands (16) supported the earlier observation from FRET (19) that dsRNA alone is sufficient of releasing CARDs. Therefore, to provide additional details on CARDs behavior in the context of FL RIG-I, we employed NMR techniques that facilitate the evaluation of highly dynamic, higher molecular weight (MW) proteins in solution. Advancements in the selective isotope enrichment of proteins, such as ¹H,¹³C-labeleling at the ϵ-methyl of methionines, as well as implementation of methyl-TROSY experiments, have made NMR spectroscopy a particularly powerful technique for evaluating biomolecular interactions (20). In our study, we applied selective methionine labeling of RIG-I allowing us to interrogate the solution behavior of FL RIG-I by NMR (MW > 100 kDa). We identified Met148, Met385 and Met923 as crucial residues, being able to sense RIG-I’s conformational changes on CARDs, Hel1 and CTD, respectively. A detailed analysis of the Met148, Met385 and Met923 behavior led to the conclusion that neither ATP nor ADP binding is correlated with ejection of CARD domains. Moreover, we found that even an antagonistic RNA is capable of CARD release in the presence and absence of nucleotides, further supporting FRET and cryo-EM conclusions coming from the Pyle Lab. Additionally, we were able to demonstrate that the interaction of the ATP transition state with RIG-I/RNA complex drives the formation of a promiscuous/alternative state which is, however, again not correlated with CARD dissociation. Finally, incorporation of NMR’s unique ability to detect solution behavior of isolated domains into comprehensive knowledge generated by multiple laboratories over the years, allowed us to strengthen the conclusions derived from most recent FRET and cryo-EM work, and fill in final details about nucleotide's impact on CARDs ejection.

MATERIALS AND METHODS

RNA synthesis

RNA hairpins and duplexes (R1–R6) were prepared by standard solid phase synthetic techniques employing suitably protected phosphoramidite building blocks on a MerMade12 oligosynthesizer. Diphosphate and triphosphate modified oligonucleotides were prepared on the solid support prior to RNA cleavage and deprotection based on a published procedure (43). CPG resin was washed with water and acetonitrile before treatment with 1:1 ammonia hydroxide:methyl amine for cleavage/deprotection. RNA silyl ethers were deprotected with 1 M TBAF in THF and the mixture was quenched with ammonia hydroxide. This solution was dried in a Genevac system without heating for 5 h. The resulting viscous material was diluted with water and washed with water four times by spin dialysis with a 3000 MWCO membrane. The concentrated material was collected and diluted with a small amount of water for purification. The desired products were purified by preparative ion pairing reverse phase HPLC.

Protein expression and purification

Wild type, M51A, M148A, M149A, M385A, M673A and M923A mutants of FL RIG-I (2–925) were expressed with an N-terminal 8His-Avi-GS tag, after cloning into a pBAC1 insect cell expression vector. The SF21 cells were adapted to SF-4 medium (BioConcept Ltd,9–34S38-I) with yeast extract and without methionine for 1 day. Before transfection the SF21 cells (0.7 × 106 cells/ml, 650 ml total volume) were spun down for 10 min at 1000 × g and immediately resuspended in SF-4 medium with yeast extract and L-methionine (methyl-13C, CLM-206-0), to a final concentration of 1 g/l. The recombinant baculovirus was added to 5 × 1 l cultures with 1.4 × 10⁶ cells/ml at MOI of 2, and incubated shaking at 125 rpm for 72 h at 27°C. Finally, cells were collected by centrifugation and stored at −80°C.

Both wild type and mutant proteins were purified as described below. Frozen pellet was resuspended in 50 mM TRIS, pH 8.0 and 300 mM NaCl and disrupted by microfluidizer in the presence of the Roche cOmplete™ Protease Inhibitor tablets, without EDTA. The cell extract was centrifuged twice, and supernatant was loaded onto HisTrap FF Crude column. Once loading was completed resin was washed with 50 mM Tris, pH 8.0, 300 mM NaCl, 20 mM imidazole and protein was eluted upon gradient increase to imidazole concentration of 500 mM, during 10 column volumes. After over-night TEV digest the RIG-I was separated from the N-terminal 8His-Avi-GS tag and TEV protease by a second affinity chromatography step and followed by HiTrap Heparin HP column, with the elution buffer containing 50 mM Tris, pH 8.0, 1M NaCl. During the final step of purification, the protein was subjected to size exclusion chromatography in 25 mM HEPES, pH 7.4, 200 mM NaCl buffer with 5% glycerol. SDS-PAGE gels illustrating the expression and purification steps for wild type and mutant RIG-I production are included in the Supplementary Data (Supplementary Figures S12–S14).

NMR experiments

Protein samples were exchanged into deuterated buffer composed of 20 mM PBS, pD-7.1, 150 mM NaCl, 5 mM DTT and concentrated to 40 μM. The RNAs were all stored in aqueous solution, then prior to sample preparation annealed for 5 min at 95°C. The samples without nucleotides were prepared in 96 Nunc conical bottom RNAse free plates with RNAs added to each sample at a final concentration of 60–80 μM and incubated at room temperature on IKA MTS 2/4 at 300rpm shaker for 30min. For the experiments with ATP and ADP analogs samples were prepared in 96 well plates with the AMP-PCP (Sigma-Aldrich, M7510), AMP-CP (Sigma-Aldrich, M3763) and MgCl₂ added at 0.25 mM, 0.4 mM and 2.5 mM respectively, preincubated for 30min with protein, and where needed followed by RNA addition and pre-incubation at room temperature on IKA MTS 2/4 at 300 rpm shaker for 30 min. Total sample volume of 140 μl was transferred to 3 mm NMR tube and ¹H–¹³C HMQC experiments were acquired using the Bruker hmqcphpr pulse sequence at 22°C on an 800 MHz Bruker Avance III HD spectrometer, equipped with a 5 mm TCI H&F C/N-D cryogenic probe. The relaxation delay was set to 1.5 s and the number of transients was defined as eighty. The datasets were acquired with 2048 data points in the ¹H dimension and 192 increments in the ¹³C dimension, typically data collection was completed in ∼6 h for each spectrum. The spectra widths were set to 10 ppm in the ¹H dimension (centered at 4.7 ppm) and 30 ppm in the ¹³C dimension (centered at 11 ppm). Each experiment was referenced to the chemical shift of TSP (internal standard δ 0.0 ppm). Spectra were processed with Topspin 3.5pl7 (Bruker) and analyzed using NMRViewJ software. Data in Figures 3–5, Supplementary Figures S1, S3, S4, S5, S6, S7, S10 and S11 were apodized with a QSINE window function prior to Fourier-transformation. Data in Figures 2 and S2 were apodized with a QSINE window function and forward linear-predicted (8 LP coefficients) was implemented prior to Fourier-transformation.

Figure 3.

(A) Cartoon representation of domain rearrangements in FL RIG-I as a result of (R3) dsRNA hairpin loop binding, leading to release of the CARDs. Assigned Met residues on CARDs are schematically highlighted as pink rectangles, while Met residues on the remaining parts of the protein that are not under discussion are grayed out. (B) Overlay of ¹H–¹³C HMQC NMR spectra of apo (green) and R3 bound (pink) uniformly Met-[ϵ ¹³CH₃] labeled FL RIG-I indicating pronounced chemical shift changes and linewidth sharpening of assigned CARD residues upon RNA binding (M51, M148 and M149), suggesting release of CARDs. In addition, the significant chemical shift change of M385 is consistent with conformational rearrangement of the Hel core.

Figure 4.

Binding of ADP-AlF_x to binary FL RIG-I/RNA complexes leads to conformational rearrangements of regions within the CTD and helicase core, while not further affecting CARDs, which become already completely ejected into solution upon dsRNA binding. The ¹H–¹³C HMQC NMR spectra for FL RIG-I in complex with R1 (blue) and R3 (red) before A), and after B) addition of transition state analog, ADP-AlF_x. Assigned Met residues on CARDs are schematically highlighted as pink rectangles, while Met residues on the remaining parts of the protein that are not under discussion are grayed out.

Figure 5.

Non-functional as well as antagonistic dsRNAs are capable of CARDs release in binary complexes, while helicase compaction only occurs when ADP-AlF_x is bound. (A) Overlay of ¹H-¹³C HMQC NMR spectra for FL RIG-I in complex with either R4 + ADP-AlF_x (orange) or R5 + ADP-AlF_x (purple); (B) overlay of ¹H–¹³C HMQC NMR spectra between apo FL RIG-I (teal) bound to smallest functional region of lnc-Lsm3b RNA RIG-I antagonist (orange). Assigned Met residues are schematically highlighted as rectangles using the color scheme matching Figure 1B.

Figure 2.

Characteristic changes in the ¹H chemical shift of Met923-[ϵ ¹³CH₃] within the CTD of FL RIG-I upon interaction with dsRNA differentially modified at 5′ end. (A) Zoom overlays into the Met923 region of ¹H–¹³C HMQC NMR spectra for FL RIG-I in complex with: (A) either R1 (green) or R3 (pink), revealing in grey an 0.013 ppm ¹H downfield shift of Met923 between 5′-OH and 5′-ppp dsRNA hair-pin loop; (B) either R4 (peach) or R5 (teal), highlighting in grey an 0.03 ppm ¹H downfield shift of Met923 between 5′-OH and 5′-ppp dsRNA duplex; (C, D) either R1 (green), R2 (blue) or R3 (pink) illustrating in grey an 0.015 ppm ¹H downfield shift of Met923 between 5′-OH and 5′-pp/5′-ppp dsRNA hairpins.

Residue assignments were achieved (Supplementary Figures S1-S4) using the following samples: ¹³CH₃-methionine labeled FL RIG-I WT versus FL RIG-I M51A ± RNA and nucleotides, FL RIG-I WT versus FL RIG-I M148A ± RNA and nucleotides, FL RIG-I WT versus FL RIG-I M149A ± RNA and nucleotides, FL RIG-I WT versus FL RIG-I M385A ± RNA and nucleotides, FL RIG-I WT versus FL RIG-I M673A ± RNA and nucleotides, and FL RIG-I WT versus FL RIG-I M923A ± RNA and nucleotides.

For the samples prepared in presence of ADP-AlF_x, the protein was concentrated to 80 μM. The ADP-AlF_x was freshly prepared through successive addition of 4 mM MgCl₂, 20 mM NaF, 4 mM AlCl₃ and 4 mM ADP, resulting in final concentration of 4 mM ADP-AlF_x. Final samples were made through combining protein and ADP- AlF_x in 1:1 ratio, yielding sample with 40 μM RIG-I, 2 mM ADP-AlF_x and 60–80 μM of RNA.

Cell based assay

RIG-I agonist RNAs were prepared in phosphate-buffered saline (PBS) and added to 384 tissue culture plates, at 5 μl per well. LyoVec™ (InvivoGen) was added at 0.078 mg/ml to the wells containing RNA, at 5 μl per well. A549-Dual Luciferase reporter cells or A549-Dual Luciferase RIG-I knockout reporter cells (InvivoGen) were thawed from cryopreservation, diluted 5-fold in growth medium (DMEM + 10% fetal bovine serum), and centrifuged at 200 × g for 5 min at room temperature. Supernatant was removed, and cells were resuspended at 2 × 10⁶ cells/ml in growth media. 20 μl of cells were plated on top of the RNA/LyoVec™ mixtures. Assay plates were incubated for 18 h in a humidified chamber at 37°C and 5% carbon dioxide. After 18 h of incubation, assay plates were centrifuged for 5 min at 200 × g. 10 μl supernatant was removed from each well and transferred to a 384 well AlphaPlate (PerkinElmer). QUANTI-Luc™ reagent (InvivoGen), dissolved to 1x in deionized water, was added to the wells at 50 μl and mixed thoroughly. Plates were incubated at room temperature for 45 min. Luminescence values were read on an Envision multi-label reader (PerkinElmer).

Protein thermal stability experiments

NanoDSF experiments were performed using the Prometheus NT.48. Purified FL-WT RIG-I was diluted into 20 mM PBS, pH 7.5 to reach a final concentration of 2.5 μM. The RNAs were added into each sample at a final concentration of 20 μM with 100 μM ATP or ADP analogs. Prior to thermal stability test the solutions were incubated for 15 min at room temperature. Each sample was loaded into standard treated capillaries and thermal unfolding was monitored over a temperature range from 20 to 90°C, with a heating rate of 1°C/min. Fluorescence at 330 and 350 nm were detected in parallel.

Microscale thermophoresis experiments

The interactions between FL RIG-I and RNAs were determined using microscale thermophoresis Monolith NT.LabelFree system. For each sample 10 nM FL RIG-I was mixed with varying concentrations of RNAs, two-fold dilutions from 5 μM to 0.16 nM, in 20 mM Na₂PO₄, 100 mM NaCl, 5 mM DTT, 0.05% PluronicF217 pH-7.1 DTT. For the experiments with ATP and ADP analogs, the AMP-PCP, AMP-CP and MgCl₂ were added at 0.25, 0.4 and 2.5 mM, respectively. The reaction mixtures were loaded into the Monolith NT LabelFree Premium Capillaries pre-incubated to 25°C and acquired using 20% LED and 40% MST power. Data analysis was performed using NanoTemper MO. Affinity Analysis software using standard K_D fitting model based on the law of mass action.

RESULTS

The affinity of RNA matched pairs for RIG-I correlates with their ability to upregulate the interferon regulatory factor pathway

We generated five dsRNA sequences: 5′-OH, 5′-pp, and 5′-ppp analogs of a 10-base pair RNA hairpin loop (R1, R2 and R3), and 5′-OH and 5′-ppp analogs of 19-base pair dsRNA duplex (R4 and R5) (Figure 1A). The above oligonucleotides were selected as an optimal combination of RNAs that can probe molecular patterns responsible for inducing RIG-I agonist mediated signaling. We used an A549-Dual Luciferase reporter assay to test the ability for each RNA to trigger the interferon regulatory factor (IRF) pathway through RIG-I agonism. The measured EC₅₀ values confirmed that the addition of both 5′-ppp and 5′-pp modification converts weakly active RNA ligands into highly active RIG-I agonists (Table 1).

Figure 1.

(A) Structural representation of R1–R5 RNAs used in this work, bearing different 5′ modifications. (B) Schematic representation of FL RIG-I domain architecture and its spatial arrangement in the autorepressed state with NMR assigned Met residues annotated. Domains are colored as follows: CARD1 and CARD2 (pink), Hel1 (purple), Hel2i (yellow), Hel2 (green), bridging domain (black), and CTD (grey).

Table 1.

Hairpin and duplex RNAs with corresponding 5′ modifications and cellular EC₅₀ values. EC₅₀ values were generated by treating A549-Dual Luc cells with a titration of the indicated RNA in complex with a transfection reagent. 5′-OH = unmodified RNA; 5′-pp = diphosphate RNA; 5′-ppp = triphosphate RNA

Hairpin RNA	EC50 (nM)	Duplex RNA	EC50 (nM)
5′-OH (R1)	163	5′-OH (R4)	>5000
5′-pp (R2)	6
5′-ppp (R3)	4	5′-ppp (R5)	18

We also measured equilibrium dissociation constants (K_D) for these selected RNAs in the context of FL RIG-I to evaluate if there is a correlation between their activity and affinity. We used the label-free microscale thermophoresis (MST) approach that relies on detecting a change in fluorescence induced by changes in molecular motions along microscopic temperature gradients. This can be correlated to changes in hydration shell, charge, or size when a protein transitions from the free to its bound state upon ligand binding. To test the method's sensitivity, we first acquired data for R1 and R3, for which prior literature data were available. The resulting data for R1 with K_D = 5.5 ± 2.4nM were in very good agreement with previously observed values of K_D = 4.29 nM (Table 2) (21). However, when testing R3, we observed that our assay's sensitivity bottomed out around 1 nM, and a discrepancy between our measured K_D = 2.8 ± 1.6 nM and the K_D = 0.45 nM reported in the literature was observed. As a result, we concentrated our follow-up experiments on the RNAs with an affinity for RIG-I ≥ 2.5 nM. Despite this limitation, we observed that binding of 5′-ppp RNA results in higher affinity interaction with RIG-I than the corresponding 5′-OH analog. Finally, a clear trend was observed for both pairs of 5′-OH and 5′-ppp RNAs between their affinity for FL RIG-I and the extend to which they can stimulate the IFN pathway, confirming that for analyzed RNAs, the presence of 5′-ppp RNA modification is correlated with RIG-I’s enhanced ability to signal (Tables 1 and 2).

Table 2.

Equilibrium dissociation constants (K_D) for each listed RNA with FL RIG-I determined using microscale thermophoresis (MST) in either the binary complex or the corresponding nucleotide ternary complexes

	No nucleotide (observed)	No nucleotide (literature)	With AMP-CP	With AMP-PCP	With ADP-AlFx
Hairpin RNA	K D (nM)	K D (nM)	K D (nM)	K D (nM)	K D (nM)
5′-OH (R1)	5.5 ± 2.4	4.29 (21)	12.3 ± 3.2	16 ± 4.5	3.9 ± 1.9
5′-ppp (R3)	2.8 ± 1.6	0.45 (21)	2.5 ± 1.3	2.7 ± 1.4	3.5 ± 2.4
Duplex RNA	K D (nM)		K D (nM)	K D (nM)	K D (nM)
5′-OH (R4)	210 ± 76	NA	121 ± 50	220 ± 97	11 ± 8
5′-ppp (R5)	1.3 ± 0.5	NA	1.2 ± 0.3	2.9 ± 1.3	1.6 ± 0.9

Determination of a suitable labeling scheme for FL RIG-I and NMR resonance assignment

Combining methyl TROSY experiments with site-specific ¹H–¹³C methyl labeling in a deuterated background enables the study of structure and dynamics of high molecular weight proteins and their complexes in some molecular detail. However, this labeling approach requires protein expression in a bacterial system, often non-compatible with mammalian proteins. With that in mind, we first attempted production of one of the truncated RIG-I constructs (residues 230–925, Figure 1B), using the [U-²H], Ile-[δ₁–¹³CH₃], Leu-[δ-¹³CH₃], Val-[γ-¹³CH₃], labeling scheme. However, even after expression optimization we obtained <0.1 mg of protein from 5 l of Escherichia coli culture. Consequently, we focused on incorporating ¹³CH₃-methionine into protonated RIG-I in insect cells, enabling us to successfully produce the desired protein and maximize spectral resolution and sensitivity in a selected region of the NMR spectrum, which otherwise is significantly hampered in high molecular weight systems. Next we strived to generate full-length RIG-I (FL RIG-I composed of residues 2–925, Figure 1B), knowing that it will be crucial to address questions around RIG-I activation using the intact protein. We counted 11 cross-peaks, corresponding to 68% of methionines in the primary sequence, with multiple overlapped cross-peaks from flexible loop regions dominating the center of the ¹H–¹³C HMQC NMR spectrum (Supplementary Figure S1a). To identify crucial resonances representative of residues on CARDs, Hel and CTD domains, we produced methionine to alanine FL RIG-I mutants and superimposed ¹H–¹³C HMQC NMR spectra between apo wild type and each mutant, thereby revealing localization of the peaks for Met51 (CARD), Met148 (CARD), Met385 (Hel1) and Met923 (CTD) (Supplementary Figure S1b). We were unable to assign Met149 (CARD) in the WT protein due to peak shifts in the NMR spectra for more than one residue. Residues Met51, Met673 and Met923 are present in the central region of the spectra, which is dominated by methionines that are localized within flexible regions of FL RIG-I. While peaks for Met51 and Met923 were resolved enough to be included in the further analysis, the overlap of Met673 was too severe to use it as probe of RIG-I behavior. Careful inspection of the ¹H–¹³C HMQC NMR spectra, when bound to different types of RNA, enabled us to unambiguously assign the cross-peaks belonging to Met148, Met149, Met385 and Met923 in the RNA bound form (Supplementary Figures S2–S4).

The CTD is the key specificity element, enabling FL RIG-I to differentiate between RNA’s 5′-end modifications

A comprehensive recent study revealed the physical basis for how RIG-I specifically recognizes the unique molecular features of viral RNA ligands and selectively distinguishes them from closely related RNAs abundant in host cells (16). Through a series of high-resolution cryo-EM structures on FL RIG-I dsRNA complexes these authors found that FL RIG-I uses two different conformations to recognize host and viral ligand classes and that interactions with CTD are key in distinguishing host from viral RNA. These findings are consistent with the earlier observation that one phosphate residue makes the difference between an activating (5′-ppp or 5′-pp) and antagonistic (5′-p or 5′-OH) RNAs (22). Therefore, to validate our system we first evaluated if methyl NMR spectroscopy of FL RIG-I can provide insights into its interaction with 5′-OH and 5′-ppp RNA and possibly identify NMR spectral signatures that correlate with a signaling-active conformation. Interestingly, pairwise comparison of ¹H–¹³C HMQC NMR spectra for 5′-OH (R1) and 5′-ppp (R3) RNA hairpin loop bound FL RIG-I revealed an upfield shift of one of the methionine peaks, reflecting the loss of 5′-ppp modification (Figure 2A). Careful inspection of NMR spectra for several methionine to alanine FL RIG-I mutants, when bound to different types of RNA, enabled us to unambiguously assign that cross-peak to Met923 (Supplementary Figures S2a, S2b). To examine whether perturbation of the Met923 peak can be more broadly attributed to changes induced by 5′-RNA end modification, we analyzed an additional pair of 5′-ppp and 5′-OH dsRNA duplexes (Figure 2B). In fact, a general pattern emerged in which the Met923 peak in ¹H–¹³C HMQC NMR spectra of 5′-OH RNA/RIG-I complex is localized on average ∼0.02 ppm upfield from that of the corresponding 5′-ppp RNA/RIG-I complex (Figure 2A, B). These NMR results suggest that Met923 localized on the surface exposed side of CTD can sense crucial interactions of RNA with FL RIG-I that drive its specificity. Additionally, a larger shift of Met923 peak in the R4/R5 RNA pair than in the R1/R3 pair may suggest a more pronounced change in RNA affinity when 5′-ppp modification is absent. Our results are in agreement with previously published data for 5′-ppp as well as 5′-OH RNA in the context of FL RIG-I, ΔCARD, and isolated CTD, where measured values of equilibrium dissociation constants (K_D) showed that the CTD is the main specificity element, driving key affinity interactions with RNA (23).

Knowing that 5′-ppp and 5′-pp RNAs can effectively activate RIG-I, we decided to investigate if the presence of the γ-phosphate can lead to differential interactions with the CTD, sensed by Met923. Superimposition of ¹H–¹³C HMQC NMR spectra for 5′-pp and 5′-ppp RNA bound to FL RIG-I (Figure 2C) revealed that the cross-peak for Met923 has similar chemical shifts in both complexes and differentiates them from 5′-OH RNA (Figure 2D). These results led us to conclude that the interactions influencing the chemical shift of Met923 in the context of FL RIG-I are conserved between 5′-pp and 5′-ppp dsRNA. Altogether, the ligand-induced changes in the Met923 peak are in full agreement with the in-house affinity data (Table 2), as well as previously published binding data where 5′-ppp and 5′-pp RNA bound with comparable affinity to FL RIG-I (22,24). We can conclude that Met923, localized within the CTD, is a perfect remote probe capable of sensing RNAs bearing different 5′-end modifications. This work further supports the hypothesis that RIG-I interaction with the γ-phosphate is not crucial in recognition of pathogenic RNA and aligns with conclusions made for 5′-OH, 5′-p, 5′-pp and 5′-ppp RNAs based on cryo-EM and careful mechanistic studies (16,22).

RNA binding to RIG-I is sufficient for ejection of the CARDs

Multiple crystallographic studies have revealed pronounced rearrangements in truncated constructs of RIG-I (composed of Hel1, Hel2i, Hel2, Br and CTD domains) under binary (RIG-I with RNA) and ternary complex conditions (RIG-I with RNA and ATP analogs). Most recent cryo-EM work described in detail structural characteristics of agonistic and antagonistic RNA in complex with FL RIG-I. However, we are still lacking a direct structural visualization of each individual CARD in binary and ternary complexes of FL RIG-I. Zheng et al. utilized HDX-MS to establish that dsRNA binding causes the closing of helicase and CTD domains around dsRNA, which loosens the interaction between CARD and Hel2i and may require ATP binding to fully release the CARDs (17). The SAXS work by Shah et al. also proposed that CARD release is correlated with ATP binding (18). However, subsequent FRET experiments contradicted those earlier results indicating that ATP has little effect on CARD ejection (19). To better understand structural rearrangements of CARDs in FL RIG-I at the molecular level, we turned to NMR studies of uniformly Met-[ϵ-¹³CH₃] labelled FL RIG-I. The chemical shifts of Met51 and Met148 were used as reporters of CARD domain apo structure, while Met148 and Met149 served as probes of changes induced in the CARD region upon RNA binding. Comparison of ¹H-¹³C HMQC NMR spectra of FL RIG-I bound to 5′-OH or 5′-ppp analogs of an RNA hairpin loop (R1 and R3) with the spectrum of apo FL RIG-I showed that the methyl resonances within the CARD domains undergo pronounced chemical shift perturbations in conjunction with linewidth sharpening upon RNA binding, suggesting the release of CARD domains (Figures 3A, B, Supplementary Figures S5 and S6). Moreover, comparing the spectra for both RNA/RIG-I binary complexes with 5′-OH and 5′-ppp analogs of RNA hairpin loops R1 and R3 showed very similar Met148 and Met149 regions, demonstrating that CARD release is not dependent on the presence of 5′-ppp modification (Supplementary Figure S6). Interestingly, the resonance for Met148, which is clearly visible in a well-dispersed region of the NMR spectrum, shows very pronounced line broadening in apo FL RIG-I, consistent with CARD domains being buried in the helicase core and tumbling at a slower rate expected for ∼100 kDa protein (Supplementary Figure S5a). In contrast, Met148 shows much sharper linewidths in the binary complexes (Supplementary Figure S5b), supporting the notion that RNA binding causes CARDs release into the solution, thus adapting relaxation properties of a small ∼20 kDa CARD subunit linked to the helicase core through an about 50-residue flexible linker. This resulted in a tumbling time that is up to ten times faster than in the case of the more compacted apo state of the FL protein (25).

Knowing that ATP binding and hydrolysis were proposed to be correlated with CARD release, we carried out additional NMR experiments to search for any chemical shift changes in the Met148 and Met149 regions that could be induced by the presence of either ATP or ADP analogs. However, the data showed no further changes in the chemical shift of Met148 and Met149 upon addition of AMP-CP-Mg²⁺ (ADP analog), suggesting that CARD domains ejection from the FL RIG-I is not affected by the presence of the hydrolysis reaction product (Supplementary Figure S7). Likewise, the addition of the ATP analog (AMP-PCP-Mg²⁺) gave the same results to those observed for the ADP analog, namely, superimposition of ¹H–¹³C HMQC NMR spectra for dsRNA hairpin loop-bound FL RIG-I in the presence of AMP-PCP-Mg²⁺. This indicated that Met148 and Met149 could not differentiate between the nucleotide-free and ATP-bound conformations of the CARD domains (Supplementary Figure S7). With those data in hand, we sought an alternative technique to further probe whether binding of the analyzed nucleotides influences RIG-I’s ability to eject CARDs in the presence of 5′-ppp and 5′-OH RNA. Differential scanning fluorimetry (DSF), which was originally developed to assess the stability or ligand-binding properties for proteins, proved to be an ideal orthogonal tool to measure changes in the stability of RNA/RIG-I complex upon interaction with ATP or ADP (26). Comparison of the transition temperatures (T_m) for the apo RIG-I and its 5′-ppp RNA complex before and after addition of ADP or ATP analog showed no significant changes in T_m values (Table 3, Supplementary Figure S8a). While there was an apparent change in T_m values when comparing the DSF data between FL RIG-I in the presence of either 5′-ppp or 5′-OH dsRNA hairpin loop (R1 versus R3 in Supplementary Figure S8b), no further significant influence on the stability of these dsRNA/RIG-I complexes could be detected by the addition of either an ADP or ATP analog (Supplementary Figures S8a, S8c). These results are fully consistent with the NMR conclusions that neither ATP nor ADP binding further influence RIG-I’s ability to release the CARDs once the complex with dsRNA is formed, supporting the conclusions derived in the Pyle Lab based on the FRET analysis of RNA interaction with FL RIG-I and various nucleotides.

Table 3.

Effects of RNA and nucleotide binding on the thermal stability of FL RIG-I are reflected by dramatic changes in the inflection point in DSF measurements relative to apo FL RIG-I (T_m = 46.6°C)

	No nucleotide	With ADP	With AMP-PCP	With ADP-AlFx
Hairpin RNA	T m in °C	T m in °C	T m in °C	T m in °C
5′-OH (R1)	53.7	53.6	53.5	66.1
5′-ppp (R3)	57.7	57.7	57.5	63.9
Duplex RNA	T m in °C	T m in °C	T m in °C	T m in °C
5′-OH (R4)	58.5	58.2	58.3	67.6
5′-ppp (R5)	59.6	60.4	59.5	64.8

The helicase domains of FL RIG-I undergo conformational changes upon RNA binding

Based on our results, we reasoned that differences in the ¹H–¹³C HMQC NMR spectra of uniformly Met-[ϵ-¹³CH₃] labelled FL RIG-I could also potentially inform on conformational changes of the helicase core that are induced by RNA binding. We strategically targeted two methionines for this purpose, namely Met673 within the Hel2 domain which is localized in the proximity of the ATP-binding site, and Met385, which is localized at the interface of Hel1 domain and the RNA (Figure 1B). Unfortunately, it was impossible to obtain an assignment for Met673 due to significant overlap in the central part of the ¹H-¹³C HMQC spectra, where the peak of Met673 is likely located. In contrast, the peak for Met385 could be easily identified in the ¹H–¹³C HMQC NMR spectra for apo RIG-I and all its RNA-bound complexes. Overlays of apo and RNA-bound FL RIG-I ¹H–¹³C HMQC NMR spectra revealed significant chemical shift perturbations of Met385 for all RNAs studied, consistent with large conformational rearrangements of the Hel core upon RNA binding (Figures 3, Supplementary Figures S4 and S5). Interestingly, in contrast to Met148 within CARD2 (vide supra), Met385 showed very pronounced line broadening in addition to significant chemical shift perturbations upon RIG-I/ RNA complex formation. Importantly, no significant changes were observed in the Met385 region of ¹H–¹³C HMQC NMR spectra between binary (RNA only) and ternary (RNA with nonhydrolyzable ATP or ADP analog) complexes of RIG-I (Supplementary Figure S7), like what we had observed for peaks from residues in the CARD domain (vide supra). This observation is consistent with data from the Pyle laboratory and suggests that the main structural rearrangements in RIG-I are primarily driven by RNA binding (19). We should mention, however, that because of the sparse NMR labeling scheme employed and the severe spectral overlap in the central region of the NMR spectra, we might miss subtle changes in the helicase core that could be caused by direct nucleotide-binding events; therefore, we refer the reader to the most recent cryo-EM studies highlighting intricacies of ATP interactions with RIG-I in the context of not only short, but also long RNAs (16).

ATP transition state analog binding to RIG-I/RNA complex drives formation of promiscuous/alternate states, uncorrelated with CARD release

Shah et al. proposed that formation of the ternary RNA/RIG-I complex with ADP-AlF_x, considered an accurate transition state analog of ATP (27), leads to compaction of helicase domains and more complete CARD release (18). An alternative analysis of RIG-I’s ternary complexes by FRET showed, however, no significant changes in CARD ejection when ATP or ADP analogs were used; although, a subtle decrease in FRET signal was observed in the context of ADP-AlF_x which was interpreted as a small change in the position of the CARDs (19). To probe whether ADP-AlF_x binding indeed leads to more complete CARD ejection in RIG-I, we analyzed changes in the RNAs’ affinity and Met-[ϵ-¹³CH₃] chemical shift perturbations upon ternary complex formation. During binding to the RIG-I/ADP-AlF_x complex, the 5′-OH and 5′-ppp analogs of RNA hairpins R1 and R3 showed dissociation constants that were comparable to values obtained for the nucleotide-free binary complexes (K_D = 5.5 nM versus 3.9 nM for R1 and 2.8 nM versus 3.5 nM for R3, Table 2). Interestingly, a comparison of the ¹H–¹³C HMQC NMR spectra between the analyzed dsRNA/RIG-I binary complexes and corresponding ADP-AlF_x ternary complexes revealed significant perturbations for the numerous NMR peaks, suggesting conformational changes occurring in multiple domains as a result of ATP transition state formation. In contrast to the AMP-CP-Mg²⁺ and AMP-PCP-Mg²⁺ data, we observed a loss of Met923’s ability to sense 5′-end dsRNA modifications in full-length RIG-I with ADP-AlF_x (Figure 4). The effect of ADP-AlF_x addition was further evaluated by comparing chemical shifts for Met385 in those complexes, leading us to conclude that only significant changes in behavior of the helicase core could lead to such pronounced chemical shift perturbations (Figure 4). To further analyze whether all these conformational rearrangements result from the final step of full CARD release, we compared the chemical shifts for Met148 and Met149 in ¹H–¹³C HMQC NMR spectra between the ADP-AlF_x bound ternary and the corresponding RNA/RIG-I binary complexes. Surprisingly, neither of the residues showed additional chemical shift perturbations in the ternary complexes, indicating that the CARD domains were already in the same fully ejected conformation in the binary complexes (Figure 4). Although, the presence of ADP-AlF_x clearly induces conformational rearrangements within the CTD and the helicase core; it does not change the behavior of the CARDs, which are already completely ejected into solution upon RNA binding. In light of our data, we can propose that a slight decrease in FRET intensity, observed in the context of ADP-AlF_x bound RNA/RIG-I complexes, results from conformational rearrangement of the helicase core, since the FRET reporter was designed to measure CARDs position relative to HEL2i domain rather than further CARD ejection. Overall, NMR and FRET observations are highly consistent, supporting the conclusion that RNA alone is sufficient for full CARD ejection regardless of the presence of ATP, ADP or ADP-AlF_x.

To further confirm and expand upon these results, we evaluated the effect of ADP-AlF_x binding on the T_m of the analyzed RNA/RIG-I complexes. All ternary complexes showed a dramatic change in the inflection point of the resulting fluorescence curve, unusually spanning over a broad temperature range (Table 3, Supplementary Figure S9). This data may indicate a variety of alternative conformations as a result of complex formation in the presence of this ATP transition analog, occupying a wider range of species with varying inflection points, overall resulting in a more bell shape-like transition curve. These results suggest a loss of FL RIG-I’s selectivity for 5′-ppp over 5′-OH dsRNA and the induction of pronounced structural changes in the presence of the ATP transition state analog. Considering the NMR data, we can propose that the detected changes in the T_m of the analyzed dsRNA/RIG-I ternary complexes are predominantly driven by rearrangements of the CTD and the helicase core, as indicated by the chemical shift perturbations of Met923 and Met385 peaks in the ¹H–¹³C HMQC NMR spectra.

Evaluated RIG-I antagonistic RNAs are capable of CARDs ejection, but reveal a more complex helicase core behavior

After gaining a better understanding of the behavior of highly active agonistic RNAs in their binary and ternary complexes with RIG-I and nucleotides, we tested the effects of RNAs that either cannot activate RIG-I or are known to antagonize it. We first used a set of 5′-OH and 5′-ppp 19-base pair dsRNA duplexes (R4 and R5), in which R4 served as a negative control that is unable to activate RIG-I in vitro. Again, 5′-ppp 19-base pair dsRNA in the context of either the binary complex with RIG-I only or in ternary complexes with RIG-I and either AMP-CP-Mg²⁺, AMP-PCP-Mg²⁺ or ADP-AlF_x showed very similar affinities in all four analyzed cases by MST (K_D = 1.3, 1.2, 2.9 and 1.6 nM respectively, Table 2). These results agree with studies for agonistic R3, where no clear trend in RNA affinity towards RIG-I was observed; nevertheless, it needs to be acknowledged that we can only rank order affinities ≤2.5 nM due to limitations in the MST assay set-up. In contrast to agonistic RNA (vide supra), 5′-OH 19-base pair dsRNA duplex binding to RIG-I showed over an order of magnitude increased affinity to the ADP-AlF_x bound state of RIG-I (K_D = 210 nM in the absence of nucleotide versus 11 nM in the presence of ADP-AlF_x, Table 2). Interestingly, this behavior was not observed in the case of antagonistic R1 which showed very similar affinities in all four analyzed cases by MST (AMP-CP-Mg²⁺, AMP-PCP-Mg²⁺ or ADP-AlF_x). In the previous paragraph, we had shown by NMR that the RIG-I protein can adopt very different, more promiscuous conformations when bound to RNA hairpin duplexes in the presence of the transition state analog. To test the extent of this observation, we acquired ¹H–¹³C HMQC NMR spectra for R4 and R5 in the absence and presence of all three nucleotides. The NMR data for the 5′-ppp 19-base pair dsRNA matched observations made for the R3 RNA hairpin, namely chemical shift perturbations for Met148 and Met149 consistently indicated full CARDs release as a result of RNA binding, irrespective of RIG-I interaction with nucleotides (Supplementary Figures S7 and S10). The resonance for Met923 agreed with the chemical shift induced by the presence of 5′-ppp modification on RNA; however, the ability of Met923 to discriminate between RNA’s 5-end modifications was lost in the presence of ADP-AlF_x (Supplementary Figure S10c). Finally, the peak for Met385 confirmed the engagement of the helicase core upon RNA binding and a large rearrangement within the helicase core upon interaction with ADP-AlF_x (Supplementary Figure S10c).

Next, we acquired the same set of NMR experiments for 5′-OH 19-base pair dsRNA duplex (R4). However, in the binary complex we already witnessed an unexpected ‘apo’ peak for Met385 co-existing with the set of CARDs and CTD peaks indicating the presence of the bound conformation (protein-ligand ratio used ensured >99% of RIG-I bound form, Supplementary Figure S11a). We further tested whether this behavior is not a result of binding interaction with 1:2 stoichiometry, through generation of a shorter 5′-OH 11-base pair RNA construct (R6), consisting of an engineered single blunt end and an added hairpin loop (Supplementary Figure S11b). As in the case of full length R4 duplex, we still observed an ‘apo’ peak for Met385 co-existing with the set of CARDs and CTD peaks representative of the bound conformation (Supplementary Figure S11b). In addition, to exclude the possibility that the changes we observed were due to the presence of not annealed single stranded (ss) RNA, we acquired data in the presence of ssR4 (Supplementary Figure S15). The only residue perturbed by RNA binding was Met923, confirming ssR4 interacts only with CTD and indicating that changes we observe for dsR4 are caused by its differential binding.

The behavior of R4 is clearly unique. In the case of R1 RNA, we observed changes in the NMR spectra that agreed with CTD engagement, CARD release and most importantly helicase core compaction as in the case of matched pair 5′-ppp agonistic R3. Knowing that in the most recent work from the Pyle Lab no pronounced differences could be observed in the Met385 region between 5′-ppp, 5′-pp, 5′-p and 5′-OH dsRNAs (16), we decided to analyze the 5′-OH dsRNA published in the cryo-EM study to evaluate whether it represents more R1 or R4 like behavior. Cleary, chemical shift perturbations induced by literature 5′-OH dsRNA correlate well with R1 chemical shift pattern but not with R4 pattern (Supplementary Figure S16), making the interpretation of unique R4 data challenging in the context of RIG-I conformations identified to date.

Knowing that we observed an order of magnitude gain in affinity upon adding ADP-AlF_x, we also acquired ¹H–¹³C HMQC NMR spectra for the ternary complex of R4, RIG-I and ADP-AlF_x. Data for the R4 ternary complex were strikingly different from the binary complex and resembled more the ¹H–¹³C HMQC NMR spectra for the ternary complex, consisting of R5, RIG-I and ADP-AlF_x suggesting a major structural reorganization (Figure 5A). This resulted in a large chemical shift perturbation for Met385, suggesting that the helicase core of R4 can potentially be only fully engaged upon interaction with ADP-AlF_x. Furthermore, even though R4 cannot activate RIG-I in vitro, it was still able to fully eject the CARDs (Figure 5A).

To experimentally address the possibility that there are additional antagonistic RNAs capable of releasing the CARDs and inducing an alternative conformation of the helicase core, we also acquired a corresponding set of ¹H–¹³C HMQC NMR spectra for a smallest functional region of lnc-Lsm3b RNA, known to act as a high-affinity RIG-I antagonist (Figure 5B) (28). Indeed, data for R4 and this smallest functional region of lnc-Lsm3b RNA bound to RIG-I were nearly identical, in that Met385 (Hel1) resonated at the ‘apo’ frequency, while Met923, Met148, Met149 (CTD and CARD) clearly showed chemical shift perturbations that were characteristic of RNA binding (Figures 5B, Supplementary Figure S11b). Taken together, it seems reasonable to assume that CARD release is a highly nonspecific event for RIG-I that can be triggered by the binding of different types of RNA.

DISCUSSION

It is well established that RIG-I exists in the closed state in uninfected cells, where CARDs responsible for caspase activation and recruitment are masked to prevent nonspecific stimulation of an immune response (29). However, upon viral infection, virus-specific RNA species associate with RIG-I, leading to changes in RIG-I conformation. This conformational change leads to the steric displacement of CARDs from the inhibitory state; as a result, CARDs become poly-ubiquitinated and capable of inducing polymerization of MAVS, leading to activation of IRF3 and induction of IFN-α/β (30–32). Therefore, it is broadly accepted that tandem CARDs act as the RIG-I signaling domains, and it has been shown that even a single amino acid substitution within the first CARD leads to abolishing its function (33). In contrast, the presence of both CARDs is required to retain their function (34). Remarkably, overexpression of tandem CARDs alone is sufficient to trigger downstream signaling leading to the induction of type I IFN (35). Due to the importance of understanding the main drivers leading to the unmasking of CARD domains upon viral infection, in the present study we investigated the correlation between RNA oligonucleotide binding and the release of CARD domains into solution.

As a DExD/H helicase, RIG-I is capable of binding nucleic acids and ATP hydrolysis (36,37). It has been demonstrated that ATP binds to the Hel2 domain, and ATP hydrolysis is stimulated by RNA binding but negatively regulated by CARDs when RNA is absent (32,36,38). Based on those results, a mechanism was proposed where RNA binding stimulates ATPase activity of the helicase domain, leading to conformational rearrangement required to release CARDs from the auto-inhibited state (37). Shah et al. tested the effect of ATP and its analogs, using the SAXS method, showing that the formation of ternary complexes with ADP-AlF_x leads to further compaction of the helicase core and complete CARD release (18). However, our NMR studies clearly indicate that RNA binding alone can effectively dissociate CARDs, and additional interaction with neither of the nucleotides tested impacts this process. If nucleotide binding were essential for CARD release, large chemical shift perturbations would be expected for residues localized on the CARDs when ATP or ADP analogs are added. Remarkably, we observed no chemical shift perturbations for all the assigned methionines localized on CARDs when nucleotides, even ADP-AlF_x, were added. Consistent with that, we observed that the stability of the RIG-I/RNA complex in the presence of ATP and ADP analogs, evaluated by differential scanning fluorimetry, showed no significant changes. However, we observed a dramatic change in the DSF inflection point of the resulting fluorescence curve when the ADP-AlF_x was added to RIG-I/RNA complexes. The DSF data provide only a global readout of the RIG-I/RNA/ADP-AlF_x complex behavior. Still, in combination with NMR experiments, we can demonstrate that under these unique conditions, the change in DSF inflection point shift is directly correlated with significant rearrangement in the helicase core but does not extend to CARD release. Recently, the Pyle Lab developed a FRET approach in which they evaluated relative contributions of RNA and ATP binding to RIG-I activation (19). They proposed that RNA binding alone is sufficient to liberate CARDs, and this event doesn’t require ATP binding, as no significant change in the FRET signal was observed upon ATP binding, overall agreeing with our findings. Nevertheless, they observed a small but detectable decrease in FRET intensity upon interaction with ADP-AlF_x which was interpreted as a small change in the position of the CARDs. Considering our data, this decrease in FRET intensity results from conformational rearrangement of the helicase core, as the FRET reporter was designed to measure CARDs position relative to HEL2i domain, rather than further CARD ejection. Overall, those FRET observations are highly consistent with our data supporting the conclusion that RNA alone is sufficient for CARD ejection. Henceforth, ATP can be responsible for dissolving the signaling state of RIG-I and preventing its activation by self RNA, as proposed by Hopfner et al (39). However, based on our and the Pyle laboratory's data the model proposed by Hopfner could be edited such that CARD release is only dependent on RNA binding (39). Our conclusions about CARD release are further supported by recent single-particle cryo-EM structures probing different conformations of the helicase core in the context of FL RIG-I bound to 5′-OH, 5′-p, 5′-pp and 5′-ppp RNAs that represent different types of viral and host RNA ligands in the presence and absence of ATP (16). Given the fact that these snapshots revealed multiple conformations of the helicase core, the authors conclude that RIG-I is capable of binding all RNAs it encounters, thereby nonspecifically releasing CARDs while differentiating between different types of RNA based on unique interactions they can make with the helicase core.

Due to the critical function of CARDs in antiviral signaling, one would expect that their release from the body of the helicase is tightly regulated. Therefore, the notion that any RNA can lead to CARD ejection may seem counterintuitive. Hence, to strengthen and widen findings from literature and our laboratory, we included experiments with recently identified endogenous antagonistic lnc-Lsm3b RNA, which can compete with viral RNA and prevent downstream signaling (28). It has been established that lnc-Lsm3b has multivalent structural motifs localized along the long stem structure that drives RIG-I’s inhibition, defined as the smallest functional regions of lnc-Lsm3b RNA. In our study, we analyzed the behavior of these smallest functional regions of lnc-Lsm3b RNA and demonstrated that even this endogenous antagonistic RNA is capable of CARDs release, despite inducing differential conformation of the helicase core. These observations further suggest that the structural requirements for RNA to cause CARD ejection are very broad, and there must be an alternative mechanism in place that prevents unintentional RNA signaling. Importantly, recent discoveries from the Pyle Lab showed that RIG-I uses interaction networks between helicase core and CTD to impact formation of a long-lived signaling complex and minimize signaling from low-affinity host RNAs (16). Additionally, findings by multiple laboratories suggested that successful downstream signaling leading to induction of IFN-α/β requires not only CARD release but also their self-association in tetrameric assemblies, interaction with MAVS to form mitochondria anchored filaments, and recruitment of E3 ubiquitin ligase (40,41). In particular, it has been demonstrated that mutation of residue Lys172 on RIG-I prevents its further ubiquitination and activation. Moreover, the data showed that sequential ubiquitination of the CTD is a prerequisite for CARDs ubiquitination, and finally, ubiquitination of oligomeric CARDs increases the stability of the oligomeric complex (41–43). Parts of these observations were recently questioned; however, all the findings align with the fact that an active RIG-I signal needs to be transmitted to MAVS by CARD-CARD interaction (44). In light of the data that CARDs release is a highly non-specific event, but crucial for interaction with MAVS, the RNA’s ability to form a long-lived signaling complex becomes highly important in driving effective signaling. In summary, our work shows that the CARDs release event is triggered by binding the diverse range of self and non-self RNAs rather than ATP and is further supported by findings from other laboratories, indicating that CARD signaling depends on the residence time of RIG-I on a particular RNA ligand (16).

Data Availability

The data underlying this article are available in the article and in its online supplementary material.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA. Funding for open access charge: Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.

Conflict of interest statement. All authors are employees or former employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA, and may hold stock or stock options in Merck & Co., Inc., Rahway, NJ, USA.