RNA editing at a limited number of sites is sufficient to prevent MDA5 activation in the mouse brain

Adenosine deaminase acting on RNA 1 (ADAR1), an enzyme responsible for adenosine-to-inosine RNA editing, is composed of two isoforms: nuclear p110 and cytoplasmic p150. Deletion of Adar1 or Adar1 p150 genes in mice results in embryonic lethality with overexpression of interferon-stimulating genes (ISGs), caused by the aberrant recognition of unedited endogenous transcripts by melanoma differentiation-associated protein 5 (MDA5). However, among numerous RNA editing sites, how many RNA sites require editing, especially by ADAR1 p150, to avoid MDA5 activation and whether ADAR1 p110 contributes to this function remains elusive. In particular, ADAR1 p110 is abundant in the mouse brain where a subtle amount of ADAR1 p150 is expressed, whereas ADAR1 mutations cause Aicardi–Goutières syndrome, in which the brain is one of the most affected organs accompanied by the elevated expression of ISGs. Therefore, understanding RNA editing–mediated prevention of MDA5 activation in the brain is especially important. Here, we established Adar1 p110–specific knockout mice, in which the upregulated expression of ISGs was not observed. This result suggests that ADAR1 p150–mediated RNA editing is enough to suppress MDA5 activation. Therefore, we further created Adar1 p110/Adar2 double knockout mice to identify ADAR1 p150–mediated editing sites. This analysis demonstrated that although the elevated expression of ISGs was not observed, only less than 2% of editing sites were preserved in the brains of Adar1 p110/Adar2 double knockout mice. Of note, we found that some sites were highly edited, which was comparable to those found in wild-type mice, indicating the presence of ADAR1 p150–specific sites. These data suggest that RNA editing at a very limited sites, which is mediated by a subtle amount of ADAR1 p150, is sufficient to prevents MDA5 activation, at least in the mouse brain.

Thank you for this insightful comment. In response to this suggestion, we treated splenocytes isolated from adult Adar1 p110 -/mice with IFN-β1. This analysis has demonstrated that ADAR1 p110 expression was not detectable in Adar1 p110 -/splenocytes even when ADAR1 p150 protein expression was upregulated (S2B Fig), suggesting that ADAR1 p110 is not efficiently produced from a cryptic promoter and Adar1 p150 mRNA in mice, which might be different from the phenomena reported in human cell lines. We described this result in the Result section (Page 12, and discussed the difference by citing three papers indicated by the reviewer (Page 26, lines 396-400) in addition to the methods used for this study (Page 33, line 501 -Page 34, lines 508).
2. The authors should also assess editing in ADAR1E861A/p110-/-mice to rule out any residual editing by E861A mutant.
Thank you for this suggestion. We previously reported that no editing events were detected in Adar1 E861A/E861A Adar2 KO mice (Cruz et al, RNA, 2000), indicating that E861A mutant completely loses its editing activity (Page 6, lines 85-87). Therefore, it is anticipated that ADAR1 p150 that is expressed from one allele is the sole active ADAR1 isoform present in Adar1 E861A/p110del mice. In contrast, given that ADAR2 is present in these mutant mice, we assessed the editing of only the sites shown in Fig 9, which are likely ADAR1 p150-specific sites. As shown in the new Fig 9, the editing of these sites was preserved in the brain of Adar1 E861A/p110del mice, which further suggests that a subtle amount of ADAR1 p150 is sufficient to sustain the editing ratio of certain sites. We have added a description of this result to the main text (Page 22, lines 345-347).

It will be informative to the readers if the authors could include A->G editing in the repeat sequences (SINE). This could provide information to readers about the levels of p150 editing.
Thank you for this comment. We had already categorized all the editing sites into the repeat sequences (shown as "repetitive") and non-repeat sequences ("non-repetitive") in S2 Table.

LN 39-p is missing in p150
Thank you for reminding us of this mistake. We have fixed the indicated point (Page 3, line 39).

Answers for Reviewer #2:
The ADAR family of RNA editing enzymes are important for modulating the structure and coding properties of the mammalian transcriptome. Here the authors describe the generation and characterization of mice genetically modified to express only the p150 isoform of ADAR1 and not the p110 isoform of ADAR1 or ADAR2. Since these three proteins are the only editing-functional mammalian ADARs, these mice provide an important tool for the study of phenotypes associated with p150 and p150-specific editing sites. Other labs have shown that ADAR1 p150 activity is essential to block recognition by duplex RNA sensors (e.g.

MDA5) of the duplex RNA structure present in the transcriptome. However, a critical question that has been dogging the editing community is "What is the identity of the critical RNAs modified by ADAR1 that would otherwise by recognized by dsRNA receptors and trigger and immune response?" This study
provides key data that start to address this question. By removing the other functional ADARs from the mouse genome, these authors can identify the sites edited by ADAR1 p150. The results reported in Figure 9 described highly efficient ADAR1 p150 editing sites, particularly in the 3' UTR of the Mad2ll transcript, are very important. The conclusions of this paper are supported well by the results

presented. I have only the following minor comments.
We thank the reviewer for these comments.
1) The introduction is quite wordy with the last paragraph essentially providing a synopsis of the entire study. The authors may wish to move some of this text to the discussion.
Thank you for this comment. We have shortened the last paragraph in the Introduction from 27 sentences to 10 sentences (Page 9, lines 138-141).
2) The discussion of the truncated ADAR1 p110 found in the M249A/M249A mice is distracting and does not add to the overall impact of the paper. The authors may consider removing this entirely from the manuscript.
Thank you for pointing this out. In response, we have deleted the corresponding sentence (Page 24, lines 372-374 in the original manuscript).
3) An obvious question for these authors to address is whether edited Mad2ll 3'-UTR interacts with dsRNA sensor proteins like MDA5 differently than the unedited RNA. While I don't believe these results are essential for publication, they could substantially raise the impact of the current work.
Thank you for this insightful comment. We agree that the experiment proposed by the reviewer is crucial as a next step to know how differently edited and unedited transcripts bind to and activate MDA5. However, as described in the Discussion part (Page 27, line 419 -Page 28, line 436), although 3'UTR of Mad2l1 is a prime candidate site, we currently have to continue identification of highly edited sites and determine which editing sites are critical for preventing MDA5 activation. Therefore, we would like to perform the suggested experiment as a separate study in the future. Thank you again for this comment.

Adar1 p150 and loss of p110 does not upregulate expression of p150 and also indicates that there is no interferon induction in this model.
We thank the reviewer for these comments. In particular, the comment "A milestone effort!" is very encouraging for us.

Major Comment
The p150 is known to undergo nuclear-cytoplasmic shuffling. While the editing dat presented by the authors is consistent with only a cytoplasmic localization of p150 in the p110 null mouse, it is necessary to show this. The reasons are twofold.
Firstly, the nuclear and cytoplasmic localization of the p150 Zα mutant mouse described on BioRxiv by the Maelfait group is wildtype. Second, the non-editing role of p110 may be to promote nuclear uptake of p150 by sponging up the dsRNA that inhibits the bipartite the ADAR nuclear localization signal. In the absence of p110, dsRNA accumulation in the cytoplasm could prevent p150 nuclear uptake (see PMID: 24753571).
A low nuclear uptake may also account for the lack of editing in the brain.

Important events may be missed if p150 disease relevant edits only occur in the
nucleus. The possibility that nuclear editing by p150 is certainly consistent with the other paper the Kawahara group has on BioRxiv in which the editing ratio of ~30% of sites in AdarW197A/W197A mice was more than 10% higher than that in Adar1+/+ mice, while the editing ratio of ~5% of sites in Adar1W197A/W197A mice was more than 10% lower than that in Adar1+/+ mice. I hope the experiments to confirm the cytoplasmic localization of p150 in this model won't delay publication as the question can be quickly settled by immunofluorescence or cell fractionation approaches.
Thank you for this insightful comment. In response, we performed subcellular fractionation for splenocytes isolated from Adar1 p110 -/mice. We chose splenocytes because ADAR1 p150 is abundantly expressed in addition to the distinct expression of ADAR1 p110, whereas ADAR1 p150 is expressed at very low levels in the brain. As expected, this analysis revealed that ADAR1 p150 was predominantly detected in the cytoplasmic fraction of ADAR1 p110-deficient splenocytes (S2A Fig), indicating that loss of ADAR1 p110 does not largely affect the intracellular localization of ADAR1 p150. We shows that p150 can be highly expressed in the nucleus after interferon induction.
Thank you for this comment. We have changed the corresponding sentences in accordance with your suggestion (Page 5, line 80 -Page 6, line 19).

2.Line 126 "
This analysis demonstrated that more than 98% of all the editing sites found in wild-type mice were absent in the brain" while in line 131, it is stated that "the number of editing sites was less than 10% of that found in wild-type mice" give different percentages. Does the line 131 include editing observed in all tissue?
Thank you for this comment. In contrast to the number "98%", which was calculated based on the total editing sites, the number "10%" was the proportion based on the editing sites only in 3'UTR. In response to the comment 1 from Reviewer #2, however, we have removed these sentences to shorten the Introduction part.
3.Line 129 "indicating that ADAR1 p150 does not substantially contribute to RNA editing in the nucleus". This may only be due to a lack of nuclear import in the p110 null mouse (see comments above).
Thank you for this comment. We performed an additional experiment (S2A Fig) in response to this comment; please refer to our response to your Major Comment. We should note that although the sentence in Line 129 of the original manuscript has been removed from the Introduction part in response to the comment 1 from Reviewer #2, the similar sentence can be found in Page 26, lines 402-403 in the Discussion part.

4.I think the wording in the sentence starting on line 212 may need rewriting. "The
resulting Adar1E861A/p110del mice demonstrated no obvious abnormal phenotypes that differed from those in Adar1 p110-/-mice or the embryonic lethality found in Adar1E861A/E861A mice (Fig 2I)." instead of "that differed" do the authors mean ", an outcome different from"….
Thank you for this comment. We have modified the corresponding sentence in accordance with your suggestion (Page 14, line 222). 5.I think the wording on line 923 "the number of the sites commonly identified in two mice" may need rewriting. Do the authors mean "the number of sites in common between the two mice studied"?
Thank you for this comment. In response, we have revised the sentence to "the number of sites in common between the two mice studied" (Page 66, line 1010). 7.A better reference for the statement "In addition, another mutation is found in the p150 isoform-specific Zα domain, which indicates that the reduced RNA editing activity of ADAR1 p150 is probably a cause of AGS pathogenesis" is PMID: 31320745 where there is direct evidence that p150 mutations do cause AGS.

6.Legend to
Thank you for this comment. In response, we have cited the suggested paper (Herbert et al., Eur J Hum Genet, 2020) in this sentence (Page 29, lines 448-450).