HDX-MS reveals dysregulated checkpoints that compromise discrimination against self RNA during RIG-I mediated autoimmunity

Retinoic acid inducible gene-I (RIG-I) ensures immune surveillance of viral RNAs bearing a 5’-triphosphate (5’ppp) moiety. Mutations in RIG-I (C268F and E373A) lead to impaired ATPase activity, thereby driving hyperactive signaling associated with autoimmune diseases. Here we report, using hydrogen/deuterium exchange, mechanistic models for dysregulated RIG-I proofreading that ultimately result in the improper recognition of cellular RNAs bearing 7-methylguanosine and N1-2’-O-methylation (Cap1) on the 5’ end. Cap1-RNA compromises its ability to stabilize RIG-I helicase and blunts caspase activation and recruitment domains (CARD) partial opening by threefold. RIG-I H830A mutation restores Cap1-helicase engagement as well as CARDs partial opening event to a level comparable to that of 5’ppp. However, E373A RIG-I locks the receptor in an ATP-bound state, resulting in enhanced Cap1-helicase engagement and a sequential CARDs stimulation. C268F mutation renders a more tethered ring architecture and results in constitutive CARDs signaling in an ATP-independent manner.

The authors describe their study of RIG-I function in the different steps of RNA recognition and response. The model here is to assess purified RIG-I at different stages of RNA binding, CARD release from autorepression by the CTD, ATP hydrolysis, and ultimate signaling via a luciferase IFN-beta promoter reporter assay. The approach is innovated to apply a combination of HSX-MS and a level of functional assays to determine the effect from E373A and C268F mutations on the different actions of RIG-I in interacting with nonself (PAMP) and self RNA. The previous work from this group showed that H830 imparts restriction of 2'0 methyla RNA (cap1-2 RNA) to prevent high affinity binding of self RNA. The current now assesses how RIG-I mutations associated with specific autoimmune disorders in humans permits self RNA binding and RIG-I activity binding despite the H830 restriction. They show that ATP binding by the E373A mutant of RIG-I places RIG-I in a ATPon state to release the CARDs for signaling-shown by induction of IFN-beta luciferase and RIG-I structure analysis. But C268F simply places the CARDs in a conformation displaced from autorepression and now in a signaling-on state independent of ATPase. The study vlidates the checkpoint control of RIG-I operating through 1) RN binding affected by RNA modification (methylation/cap/5'ppp), 2) ATP occupancy and hydrolysis, and 3) CARD intramolecular interactions within the Hel and CTD to control RIG-I off-on signaling conformation. The new data here is the really the careful examination of the E373 and the C268 aa sites in the regulation of RIG-I and linkage with the autoimmune conditions of Singleton-Merten Syndrome.

Specific comments:
The structural studies using HDX-MS are important to reveal distinct conformation changes of RIG-I directed by the different tester RNA species in vitro. However, the study lacks any ex vivo or in vivo validation. Can RIG-I really bind to self RNAs under the conditions within a living cell or human in the context of WT vs E373 vs C268 mutation? No evidence for these interactions are shown beyond the highly artificial tester RNAs shown in Figure 1. The authors need to conduct expression and pull-down of RIG-I and RIG-I mutants from cultured cells at least, and then assess bound RNAs using typical recovery and sequencing approaches. If their model is correct, then the mutant vs WT RIG-I should be associated with self RNAs.
In terms of RIG-I signaling the authors show IFN-beta luc assay adata but this data set needs validation beyond a simple luc assay. Analysis of ISG mRNA or protein expression (for example IFIT, IFIT2, MX, OAS..) should be conducted to verify induction of the cellular response to RIG-I signaling and IFN induction (also a measure of actual IFN production levels using an IFN ELISA assay is typically used here to validate IFN-beta luc results).
Do the RIG-I mutants differentially impact IRF3 and NF-kB activation? The authors need to assess IRF3 phosphorylation and NF-kB activation directly. In light that the different mutants redirect RIG-I checkpoint signaling actions, the impact on these critical downstream transcription factors should be addressed. It is possible that the E373 vs C268 could impact IRF3 activaiton kinetics but not NFkB , for example, by placing RIG-I in different conformations for differential interaction with down stream signaling partners that impact IRF3 or NFkB signaling.
Do the different RIG-I mutants for a stable complex with MAVS? This component of the analyses should have been included but is not even mentioned nor discussed. When activated RIG-I will form a stable complex with the MAVS adaptor protein. This complex is essential for the induction of IFN and the actions of RIG-I. Thus, one should expect that the E373 and C268F mutants would form a constitutive complex with MAVS or a complex that is enhanced by RNA. How does each mutant impact MAVS binding, and what are the differential kinetics of this interaction in the presence or absence of the different tester RNAs?
What is the impact of each RIG-I mutant in the context of virus infection where RIG-I is expected to mediate virus recognition. Do the mutants compromise this activity or do they confer enhanced resistance against infection?
Statistics should be included for all data sets applicable. Please include p values and statistical methods used for the data and differences shown.
Discussion: this section needs to include a broader presenation of how RIG-I checkpoints impact the entire signaling cascade from RNA binding, RIG-I activation, MAVS interaction, IRF activation, and gene expression.
Reviewer #2 (Remarks to the Author): The manuscript by Zheng et al investigates the molecular details of how the Retinoic acid Inducible Gene-I (RIG-I) Receptor interacts with different RNAs and how mutations in RIG-I cause dysregulation. The role of RIG-I and other major intracellular immune receptors in autoimmune disease is important and more information is needed for how they activate and discriminate between viral RNA species and self RNA -and how mutations result in dysregulation and disease. In this work, the authors extend prior work that also made extensive use of the HDX-MS method, in which they studied the apo state of RIG-I and its binding to ATP, triphosphorylated RNA and longer duplex RNA (Zheng J, Nucleic Acid Research. 2015). That prior work provided a model for how RIG-I (and a related receptor) in the autoinhibited apo-state undergoes allosteric changes when transitioning to an active state. The present manuscript builds on this work and studies the binding of RIG-I (and mutants H830A, C268F and E373A) to different RNAs. The work provides new mechanistic insights into RIG-I regulation. Most interestingly, the results provide a model of how the specific gain-of-function mutations C268F and E373A in RIG-I cause dysregulation leading to Singleton-Merten syndrome, an autoimmune disease.
The experimental work, in particular the HDX-MS component, is very extensive and wellperformed. Furthermore, the conclusions of the work are interesting and impactful, also considering that the proposed mechanisms of RIG-I dysregulation by mutations could be relevant for other innate immunity receptors. I am positive towards publication. I have some specific concerns, in particular about the conclusions drawn from the EX1 kinetic data, that need to be addressed prior to publication.

Major comments:
Concerning the observed EX1 kinetics: Firstly, there are a few sentences detailing EX1 theory that needs to be revised/clarified: Line 203: The authors write: The CARD2 latch region (Y101-114), which is spatially locked to the HEL2i gate motif in the apoform, displays EX1 exchange behavior upon binding to PAMP RNA, suggesting this region of the receptor is conformationally heterogeneous in solution (4). I think this statement is unclear. The mere presence of EX1 does not show that the receptor conformationally heterogeneous. EX1 informs on the timescale at which structural transitions that facilitate exchange occurs in the protein. A protein undergoing EX2 could be equally conformationally heterogeneous.
Line 204: The authors write: "If an unfolding event occurs slow enough for the backbone amide hydrogens within the unfolding region to be fully exchanged, EX1 kinetics is observed with bimodal distribution". It is when the rate of the refolding or closing event (kcl) is sufficiently slow that EX1 kinetics is observed -please correct the sentence accordingly.
Line 210: The authors write: "Therefore, the region undergoing EX1 kinetics is a mixture of unfolded and folded conformers, which transitions from one state to the other via a slow and correlated exchange event.". This sentence is misleading, the transition is slow -and thus the exchange is correlated. Please rephrase.
Secondly, the authors should be more specific concerning what they can conclude and what they cannot solely based on the presence of EX1 kinetics observed for RIC-I. To me the authors lack functional evidence to support that the observed EX1 kinetics observed (and the folded and unfolded states involved) correspond to the inactive and active forms of RIG-I. I agree that it seems enticing to think so (and very interesting) but that does not necessarily make it so. Do the authors have functional data that show that the derived t1/2 values are on the same timescale as the rate of activation or enzymatic activity of RIG-I. If so, such a comparison would make their ultimate conclusions from the EX1 observations much more interesting and convincing.
For instance: Line 232: Due to the induced appearance of EX1 kinetics upon RNA binding the authors write "RNA binding by RIG-I drives CARDs module from a closed conformation to a partially opened conformation (Fig. 2a)".
Line 238: To more precisely measure the increased correlated exchange, we determined the CARDs transition rate from the inactive to the active state" The presence of EX1 kinetics and the derived t1/2 values shows that the rate of unfolding and refolding is slow. But the authors cannot strictly speaking conclude that the slow unfolding/refolding transition observed in the CARDs module is a transition from an inactive to an active state. Additional functional data to support this conclusion is needed, for instance as described above. Alternatively, this caveat needs to be clearly underlined.
Finaly on that note, the authors should investigate further if the auto-inhibited apo RIG-I and apo C268F, undergo EX1 kinetics by probing longer timescales (see comment below concerning  The EX1 kinetic regime of CARD2 latch peptide in apo RIG-I was difficult to detect as there is very little solvent exchange for the auto-inhibited domain" The Apo state exchanges so little that the presence or absence of EX1 kinetic cannot be determined based on the current data. This can likely be examined by prolonging the time course or maybe increasing the temperature. The current experiments are performed at 4°C. Also, it would appear that an unfolded population occurs at 1 hr for wt RIG-I bound to 3p8I, but not so for apo RIG-thus repeating the exchange experiment at longer times for both state would be informative and could reveal an effect of 3p8I. Finally, it is not clear to me why the apo-RIG is not simply referred to at wt RIG-I in the figure.
Line number 148-155. A reduction on 12 % is mentioned. It is not defined whether it is significant or not. The methods section does not describe how significance is determined -this is defined in the SI, but, in my opinion, this is an important parameter that should be easily accessible in the manuscript and when reviewing the data. Especially, when as all structural discussions are based on the HDX-MS data.
Minor comments: Abstract: Several sentences in the Abstract needs are very long and unclear and must be revised.
For instance: "A RIG-I residue (H830) mediates specific sensing of 5' 7-methyl guanosine and 2'Omethylated on the first base (Cap1) self RNA and is coupled with a threefold delay in Caspase Activation and Recruitment Domains (CARDs) partial unfolding event compared to that of 5'ppp RNA".
Line 172 and throughout: the authors use the term "solvent exchange". To me this is unspecific and strictly speaking incorrect as solvent per se is not exchanged. It could thus confuse a nonexpert -I recommend use of a more specific term like HDX, hydrogen exchange or deuterium uptake etc.  Line number 148-150. "The HDX data obtained for sequence overlapping peptides were consolidated to individual amino acid values using a residue averaging approach (29)". For people without former knowledge of HDX, it could sound like they have residue-resolved HDX-data everywhere. This is probably not the case. Please elaborate. Also, for all overlapping peptides for which this procedure was applied, the authors should inspect the maximal-labeled control and verify that the a similar back-exchange was observed. The Schriemer lab has observed that overlapping peptides can exhibit large differences in back-exchange rendering subtractive analysis problematic for those peptides. is not auto-inhibited and fully exposed to solvent, resulted in rapid deuterium incorporation and the latch peptide underwent EX2 exchange indicating structural homogeneity in solution". To me, this does not look like pure EX2 kinetics. The peak width at 10min and 15min is markedly larger than at 30min and 1h. Hinting at a possible mixture of kinetics (EXX). Furthermore, the apo CARD state is defined as "open", but its conformation is not similar to the "open" state described with blue in columns 4-9, as it takes the apo CARDs around 30min to be fully deuterated.
Line 439-440. "HDX MS data was calculated with the in-house developed software and corrected for back-exchange on an estimated 70% recovery." Figure 2a shows that the authors have recorded 'maximally labelled' samples. Did they normalize the deuterium incorporation to these peptide specific values or to the average back-exchange reported to be 30%. I would strongly recommend the former. Please elaborate on the procedure used and why. Response to comment 1: We apologize for our oversight by excluding the wealth of information already available on self RNAs binding by RIG-I. We have revised the manuscript and included the necessary information related to the self RNAs in the introduction section.
Specifically, this question concerns whether wild-type or mutant RIG-I can actually bind to self-RNAs in living cells or in vivo. Several groups have published the relevant findings and have proven the binding of RIG-I to self-RNAs found in cellular [1,12] and in vivo [2]. For instance, one study has performed RIG-I RNA immunoprecipitation from cell lysates of the murine splenic B-cell line and provided direct evidence that WT RIG-I recognizes several regions within NF-kB1 3' -UTR mRNA [12]. Similarly, an increased amount of RNA is reported to be co-purified from C268F and E373A RIG-I from uninfected cells compared to that of WT RIG-I [1]. Schuberth-Wagner et al. showed that cellular RNAs can activate H830A mutant RIG-I but not WT RIG-I [7]. Another published study using an in vivo murine model validated that small self-RNA fragments generated by RNase L can trigger IFNβ responses via the RIG-I signaling pathway [2]. Taken all together, these published studies address the question as to whether wild-type or mutant RIG-I can bind to self-RNAs in cellular or in vivo.
It is important to note that the RNAs selected in this study have been well-documented in various published articles and are considered as representative of validated RNA ligands for RIG-I. As such, these RNAs are optimal in vitro ligands [3][4][5][6][7]. The focus of the current study is to use these functionally validated RNA ligands to probe RIG-I-RNA interactions by HDX and gain insight into the structural mechanism of receptor activation. The observations from these biophysical studies are then correlated with findings in previously described studies focused on dysregulation of RIG-I in autoimmune disease.
2. In terms of RIG-I signaling the authors show IFN-beta luc assay adata but this data set needs validation beyond a simple luc assay. Analysis of ISG mRNA or protein expression (for example IFIT, IFIT2, MX, OAS..) should be conducted to verify induction of the cellular response to RIG-I signaling and IFN induction (also a measure of actual IFN production levels using an IFN ELISA assay is typically used here to validate IFN-beta luc results).
Response to comment 2: Several published studies have shown that IFN assays correlate with respective ISG mRNA expression level as well as protein levels [13,14]. For instance, one study has shown that high level of ISG15 gene expression in E373A and C268F RIG-I transfected HEK293T cells (non-infected) correlates with strong IFNB1 gene expression [13]. One recent review article also states that over-expression of mutant RIG-Is (E373A and C268F) sufficiently induced IRF3 phosphorylation and IRF3 dimerization. As a result, RIG-I mutations led to increased expression of IFN-β and ISG15 in untreated cells as well as in polyI:C transfected cells [10]. We have revised the manuscript and added this information in the introduction section.
Do the RIG-I mutants differentially impact IRF3 and NF-kB activation? The authors need to assess IRF3 phosphorylation and NF-kB activation directly. In light that the different mutants redirect RIG-I checkpoint signaling actions, the impact on these critical downstream transcription factors should be addressed. It is possible that the E373 vs C268 could impact IRF3 activation kinetics but not NFkB, for example, by placing RIG-I in different conformations for differential interaction with downstream signaling partners that impact IRF3 or NFkB signaling.

Response to comment 3:
As stated in response 2, one published study has already shown that RIG-I SMS mutants E373A and C268F similarly impact IRF3 phosphorylation activity and NF-kB activation in the non-infected HEK293T cells [13]. The impact on RIG-I downstream signaling pathways has been well described in the literature and supports the findings from our biophysical mechanistic study. Our current study presented here focuses on the direct biophysical analysis of RIG-I-RNA binding at the initial step of activation rather than related downstream signaling pathways. We have revised the manuscript and added this information in the introduction section.  [15][16][17][18][19]. A recent review article on RIG-I-like receptor states that "the importance of excess of RLR-dependent signaling via MAVS leading to IFN signature in the pathogenesis of these autoimmunity has been clarified" [9]. The barrier to RIG-I-MAVS CARD activation is the sequestration of CARDs by the Hel2i domain and involvement of K63-linked ubiquitin chains (K63-Ub n ). The Sun Hur group has solved the atomic structure of both the RIG-I CARDs tetramer attached with K63-Ub 2 and the RIG-I CARDs tetramer complexed with four MAVS CARD molecules [16,19]. These studies clearly demonstrate that RIG-I CARDs, by forming a helical tetrameric structure, acts as a template for the MAVS CARD filament assembly.
Regardless, our current study focuses on the initial activation steps of wt and mutant RIG-I to provide direct conformational dynamic information on RNA recognition and discrimination, ATP binding, and ATP hydrolysis. The MAVS CARD filament activation involves the interaction between RIG-I CARDs tetramer and MAVS CARD as well as the high affinity binder of K63 poly-ubiquitin chains [16]. While studying the kinetics of RIG-I-MAVS interaction is very interesting, it is outside the focus of the current study we present here.
What is the impact of each RIG-I mutant in the context of virus infection where RIG-I is expected to mediate virus recognition. Do the mutants compromise this activity or do they confer enhanced resistance against infection?
Response to comment 5: The RIG-I mutations E373A and C268F are implicated in autoimmunity and constitutive signaling of the receptor, rather than implicated in viral signaling, whereas the RIG-I mutant H830A has been reported to be able to recognize Cap1 RNAs found in viruses like yellow fever virus [7].
Statistics should be included for all data sets applicable. Please include p values and statistical methods used for the data and differences shown. Response to comment 1: We agree. Our interpretation of EX1 data needs more attention, and we need to avoid overinterpreting the observations. This specific sentence has been carefully revised as follows: "The CARD2 latch region (Y101-114), which is spatially locked to the HEL2i gate motif in the apo form, displays EX1 exchange behavior upon binding to PAMP RNA, suggesting that this region undergoes a partial unfolding event and structural transition that facilitates correlated exchange in the protein [5]".
Line 204: The authors write: "If an unfolding event occurs slow enough for the backbone amide hydrogens within the unfolding region to be fully exchanged, EX1 kinetics is observed with bimodal distribution". It is when the rate of the refolding or closing event (kcl) is sufficiently slow that EX1 kinetics is observed -please correct the sentence accordingly.
Line 210: The authors write: "Therefore, the region undergoing EX1 kinetics is a mixture of unfolded and folded conformers, which transitions from one state to the other via a slow and correlated exchange event.". This sentence is misleading, the transition is slow -and thus the exchange is correlated. Please rephrase.
Response to comment 2: We agree, and we have revised and extended this sentence as follows; "If a refolding event occurs sufficiently slow to allow complete deuterium exchange of backbone amide hydrogens within the unfolding region, then EX1 kinetics are observed [20]. Under EX1 conditions, if an opening or unfolding event involves more than one slow exchanging amide hydrogen, then deuterium exchange occurs simultaneously at these amides. Therefore, a bimodal distribution occurs via a correlated exchange pattern, in which the lower mass envelope corresponds to molecules that have not yet exchanged (not yet unfolded), and the higher mass envelope corresponds to molecules that have undergone exchange (molecules that have unfolded) [5,[20][21][22]. The region undergoing EX1 kinetics may represent a mixture of unfolded and folded conformers in the same unfolding event. In contrast, EX2 kinetics takes place if the refolding rate is much faster than the intrinsic exchange rate of the amide hydrogens, resulting in one isotopic envelope throughout the labeling time of the experiment." Secondly, the authors should be more specific concerning what they can conclude and what they cannot solely based on the presence of EX1 kinetics observed for RIC-I. To me the authors lack functional evidence to support that the observed EX1 kinetics observed (and the folded and unfolded states involved) correspond to the inactive and active forms of RIG-I. I agree that it seems enticing to think so (and very interesting) but that does not necessarily make it so. Do the authors have functional data that show that the derived t1/2 values are on the same timescale as the rate of activation or enzymatic activity of RIG-I. If so, such a comparison would make their ultimate conclusions from the EX1 observations much more interesting and convincing.

Response to comment 3:
We agree and apologize for overinterpreting the EX1 data. The To reduce ambiguities and overinterpretation of results, we selected RNA that had been described in detail in previously published papers for inclusion in our work. All these RNA ligands have been experimentally validated. 3p10l was reported as a minimal RNA duplex that binds to RIG-I in 1:1 ratio that stimulates robust ATPase activity and elicits a RIG-I mediated interferon response in cells [3,4]. 3p8l was reported as an inactive RNA ligand that still binds to RIG-I in 1:1 ratio but fails to induce RIG-I mediated interferon response in cells [4,6]. Cap1-10l was reported as a self RNA bearing m7G cap and 2'-O-methylation, a molecular signature of host mRNA. It binds to RIG-I in 1:1 ratio and the binding affinity drops by 200 folds comparing to 3p10l [3]. Below is a list of the high resolution atomic structures of these RNAs that we used in our study as well as information about the protein-RNA complex. Based on the wealth of published information we selected these RNAs to be part of our comprehensive biophysical study of RIG-I. 3p8l bound Helicase-RD (pdb: 4A2W) [6] 3p10l bound Helicase-RD (pdb: 5F9H) [3] Cap0-10l bound Helicase-RD (pdb: 5F98) [3] We selected these RNAs with graded efficacy from a full agonist (3p10l), partial agonist (Cap1-10l), and an inactive RNA (3p8l) that binds the receptor to perturb the conformational dynamics of wild-type and mutant RIG-I receptors. Based on our results, we developed the RIG-I activation model to suggest that the derived half-life values for each protein state (a total of 30 different states) listed in Supplementary Fig. 1a, b, and c correlate and is consistent with the RNA ligand mediated RIG-I agonism described previously. We have revised the relevant sections of the manuscript based on this concern.
For instance: Line 232: Due to the induced appearance of EX1 kinetics upon RNA binding the authors write "RNA binding by RIG-I drives CARDs module from a closed conformation to a partially opened conformation (Fig. 2a)".

Response to comment 4:
We have modified this sentence as follows; "This suggests that RNA binding to RIG-I allosterically triggers partial unfolding of the CARDs, and the extent of unfolding in solution is dependent on the efficacy of the specific RNA (Fig. 2a).
Line 238: To more precisely measure the increased correlated exchange, we determined the CARDs transition rate from the inactive to the active state" The presence of EX1 kinetics and the derived t1/2 values shows that the rate of unfolding and refolding is slow. But the authors cannot strictly speaking conclude that the slow unfolding/refolding transition observed in the CARDs module is a transition from an inactive to an active state. Additional functional data to support this conclusion is needed, for instance as described above. Alternatively, this caveat needs to be clearly underlined.
Response to comment 5: We address some of this concern in our response above to comment 3. However, we have carefully revised the sentence: "To more precisely measure the increased correlated exchange, we determined the CARDs transition rate from the lower MS envelope to the higher MS envelope.
Finally on that note, the authors should investigate further if the auto-inhibited apo RIG-I and apo C268F, undergo EX1 kinetics by probing longer timescales (see comment below concerning Fig. 2a). If they do, then one could suppose that these two forms of RIG-I should have some low residual activity, due to the slow build-up of the supposed open and active state, according to the theory put forward by the authors.

Response to comment 6:
As suggested by the reviewer, we have generated data with longer on-exchange time points. Specifically, we now have obtained HDX time points at 3, 5 and 7 hours for apo wildtype RIG-I and apo C268F mutant RIG-I as well as for RIG-I in complex with 3p8l RNA (Figure   A below). For both apo RIG-I and apo C268F RIG-I, the CARD2 latch peptide shows emerging EX1 kinetics at all of the extended time points (3, 5, and 7 hours). However, the emerging rate of the higher MS envelope in apo RIG-I and apo C268F RIG-I is sufficiently slow that curve fitting analysis fails to calculate the half-life of apo RIG-I and C268F RIG-I in the recorded HDX time points (see Figure A below). For apo C268F RIG-I, the overall conformational dynamics are higher than that for apo wild-type RIG-I in solution, as we observe higher deuterium incorporation in apo C268F RIG-I compared to that of apo RIG-I ( Supplementary Figure 1a    Also, it would appear that an unfolded population occurs at 1 hr for wt RIG-I bound to 3p8I, but not so for apo RIG-thus repeating the exchange experiment at longer times for both state would be informative and could reveal an effect of 3p8I. Finally, it is not clear to me why the apo-RIG is not simply referred to at wt RIG-I in the figure.
Response to comment 7: We have also generated new data at longer on-exchange time points for the RIG-I:3p8l RNA complex and calculated the half-life of the CARD2 latch peptide, which resulted in a value of approximately 5 hr (Figure A panel   Response to comment 8: We apologize for not being clear in the method section. We have revised the methods section as follows; "The intensity weighted mean m/z centroid value of each peptide envelope was calculated and subsequently converted into a percentage of deuterium incorporation. In the absence of a fully deuterated control, corrections for back-exchange were made on the basis of an estimated 70% deuterium recovery, and accounting for the known 80% deuterium content of the deuterium exchange buffer. When comparing the two samples, the perturbation %D is determined by calculating the difference between the two samples. HDX Workbench colors each peptide according to the smooth color gradient HDX perturbation key (D%) shown in each indicated figure. Differences in %D between -5% to 5% are considered non-significant and are colored gray according to the HDX perturbation key [23]. In addition, unpaired t-tests were calculated to detect statistically significant (p<0.05) differences between samples at each time point. At least one time point with a p-value less than 0.05 was present for each peptide in the data set further confirming that the difference was significant." HDX perturbation key Minor comments: Abstract: Several sentences in the Abstract needs are very long and unclear and must be revised. For instance: "A RIG-I residue (H830) mediates specific sensing of 5' 7-methyl guanosine and 2'O-methylated on the first base (Cap1) self RNA and is coupled with a threefold delay in Caspase Activation and Recruitment Domains (CARDs) partial unfolding event compared to that of 5'ppp RNA".
Response to comment 9: The abstract has been revised. Please see the updated abstract.
Line 172 and throughout: the authors use the term "solvent exchange". To me this is unspecific and strictly speaking incorrect as solvent per se is not exchanged. It could thus confuse a nonexpert -I recommend use of a more specific term like HDX, hydrogen exchange or deuterium uptake etc.
Response to comment 10: The term "solvent exchange" has been changed to read HDX or deuterium exchange. Response to comment 11: We have modified and highlighted R1 and R2 in red accordingly. Corrected.
Line number 148-150. "The HDX data obtained for sequence overlapping peptides were consolidated to individual amino acid values using a residue averaging approach (29)". For people without former knowledge of HDX, it could sound like they have residue-resolved HDXdata everywhere. This is probably not the case. Please elaborate. Also, for all overlapping peptides for which this procedure was applied, the authors should inspect the maximal-labeled control and verify that the a similar back-exchange was observed. The Schriemer lab has observed that overlapping peptides can exhibit large differences in back-exchange rendering subtractive analysis problematic for those peptides.
Response to comment 14: We agree and thank the reviewer for this comment. We have revised the manuscript as follows; "Digestion optimization for HDX studies resulted in greater than 90% sequence coverage for full-length WT RIG-I (~100kDa protein) as well as the gain-of-function RIG-I mutants H830A, E373A and C268F ( Supplementary Fig. 1 a, b and c). In the absence of fragmentation data from electron capture dissociation (ETD), residue level deuteration data were approximated by using HDX data from overlapping peptides, and consolidating these data using a residue averaging approach previously described [24] (Supplementary Fig. 1b). These data were calculated and mapped onto the PyMol structure model using HDX Workbench [23]." This single residue consolidation with smooth coloring approach was applied in our recent study wherein we looked at differential HDX experiments of 26 conformational states of VDRRXR heterodimer upon binding to different compounds, DNAs and co-activator proteins to aid in visualization of large datasets [25].
Additional comments: In our HDX experiment, we used 5 µl of protein sample and diluted it into 20 µl of deuterium containing on-exchange buffer such that the on-exchange sample (5 µl sample + 20 µl D2O) contains 80% deuterium content by volume. We have examined one maximally labelled control sample (H830A RIG-I) and used HDX Workbench to calculate the deuterium recovery (%) of all the analyzed peptides (see Figure C below). The average %D in this Dmax sample was approximately 70% with a range of 50%-80% ( Figure C). While our platform is not fully maximized to reduce deuterium loss post quench, the sample is maintained at low pH and all of the solvents, syringes, values, columns (except the protease column which is at 15°C) are maintained at 4°C within a cold box. The transfer tube from the deli frig to the mass spectrometer is as short as possible and insulated, and the ion source conditions are carefully set to balance deuterium loss with ion signal. With this set up we routinely obtain 70% Average, 50%-80% range deuterium recovery. Importantly, we have shown in a recent publication the reproducibility of this platform [26]. Regardless, given that all of the experiments are differentials run within the same day, the deuterium loss is expected to be equivalent for peptides under condition A versus those from condition B. Although 70% deuterium recovery is an estimate and the percentage may vary from peptide to peptide, we can use a single correction factor within HDX Workbench or use the values from the Dmax experiment [23]. This is a generally well-accepted approach for differential HDX experiments in which both samples are treated identically (with the same LC gradient, pH 2.4, 0 ℃). In addition, we generated Dmax data on the H830A mutant receptor. Using this new information, we directly compared the HDX data set (H830A RIG-I +/-3p10l) that was calculated using the single 70% deuterium recovery value and with that corrected using the Dmax control  To me, this does not look like pure EX2 kinetics. The peak width at 10min and 15min is markedly larger than at 30min and 1h. Hinting at a possible mixture of kinetics (EXX).
Furthermore, the apo CARD state is defined as "open", but its conformation is not similar to the "open" state described with blue in columns 4-9, as it takes the apo CARDs around 30min to be fully deuterated.

Response to comment 15:
We thank the reviewer for this comment and we agree. We have corrected the figure heading and replaced EX2 with 'absence of EX1'. In the text, the sentence has been revised as follows; 'In contrast, analysis of the CARDs (1-228) protein, where the CARDs domain cannot be auto-inhibited, resulted in significant deuterium incorporation and the same latch peptide showed absence of EX1 kinetics in the recorded HDX time points'.
Also, HDX analysis reveals that the isolated apo CARDs domain (CARDs only) is prone to significantly higher deuterium incorporation when compared to CARDs that is sequestrated by HEL2i such as in full-length apo receptors (WT, H830A, E373A and C268F) (supplementary  Response to comment 17: We have corrected this error.
Supporting Information: The HDX-MS experimental section has some cases of incorrect nomenglature, spellings and unexplained abbreviations etc. For instance, gHCL, D20 etc.
Please revise carefully and consistently.
Response to comment 18: We have corrected this error.
We thank the reviewer for this comment. As we previously stated the RIG-I CARDs MAVS CARD interaction is complicated, involving RIG-I CARDs tetramer formation (assisted by Lysin63 linked poly-ubiquitin chains) and nucleation of RIG-I CARDs and MAVS CARD complex assemble. We believe this is very important and should be part of a study focused on downstream signaling activation following CARDs opening. However, this study is outside the scope of the current manuscript. We do look forward to studying the MDA5 CARDs assemble with and without MAVS CARD, and compare with that of RIG-I CARDs-MAVS CARD complex.
Minor points.
1. Abstract Lines 38-40. The meaning of this sentence is unclear and ambiguous. Please rewrite and remove or qualify the reference to partial unfolding.
Author's Response: The term partial unfolding is a term used to describe the observation of EX1 exchange behavior. However, for clarity we have changed it to "partial opening." 2. Results, Lines 142 and 166. These subtitles appear completely disconnected from the text that follows them. Please replace with more accurate and descriptive subtitles, which together will provide a logical flow to the text. The subtitle at Line 166 should be moved to line 217 as a new subtitle, with the CARD2 latch peptide referenced explicitly instead of "CARDs partial unfolding events". A new subtitle should be written for the section lines 166-216, referencing the CTD and cap recognition.