key: cord-0314423-9plvbe2b
authors: Kleinova, Renata; Leuchtenberger, Alina F.; Giudice, Claudio Lo; Tanzer, Andrea; Derdak, Sophia; Picardi, Ernesto; Jantsch, Michael F.
title: The ADAR1 editome reveals drivers of editing-specificity for ADAR1-isoforms
date: 2021-11-24
journal: bioRxiv
DOI: 10.1101/2021.11.24.469911
sha: 98a47c72967097e253d6ae768bc24b5d1c4cd0ab
doc_id: 314423
cord_uid: 9plvbe2b

Adenosine deaminase acting on RNA (ADAR) (also known as ADAR1) promotes A-to-I conversion in double-stranded and highly structured RNAs. ADAR1 has two isoforms transcribed from different promoters: ADAR1p150, which is mainly cytoplasmic and interferon-inducible, and constitutively expressed ADAR1p110 that is primarily localized in the nucleus. Mutations in ADAR1 cause Aicardi – Goutières syndrome (AGS), a severe autoinflammatory disease in humans associated with aberrant IFN production. In mice, deletion of ADAR1 or selective knockout of the p150 isoform alone leads to embryonic lethality driven by overexpression of interferon-stimulated genes. This phenotype can be rescued by concurrent deletion of cytoplasmic dsRNA-sensor MDA5. These findings indicate that the interferon- inducible p150 isoform is indispensable and cannot be rescued by the ADAR1p110 isoform. Nevertheless, editing sites uniquely targeted by ADAR1p150 but also mechanisms of isoform- specificity remain elusive. Here we combine RIP-seq on human cells expressing ADAR1 isoforms and combine this with analysis of isoform-specific editing patterns in genetically modified mouse cells to extensively investigate ADAR1-isoform binding- and editing characteristics. Moreover, using mutated ADAR variants, we examine the effect of two unique features of ADAR1p150 on its target specificity: 1) cytoplasmic localization and 2) Z-DNA binding domain α. Our findings indicate that ZBDα contributes only minimally to p150 editing-specificity and that isoform-specific editing is directed mainly by the cytoplasmic localization of the editase.

A- 1090 

In our study, we employ a combination of RIP-seq on human cells and isoform-specific editing 134 pattern analysis in mouse cells to extensively examine ADAR1-isoform binding-and editing 135 characteristics. We collect and comprehensively analyze a high-quality ADAR1p150-and 136

ADAR1p110-specific mouse-editome. Moreover, using mutated ADAR versions, we 137 investigated the role of cytoplasmic localization and of ZBDa on ADAR1p150 editing-138 specificity. Our findings suggest that ZBDa has a minor contribution to editing specificity and 139 that ADAR1p150 editing patterns are mostly driven by the cytoplasmic localization of this 140 protein isoform. 

To determine the target specificity of ADAR1 isoforms, we employed two different 147 approaches: on the one hand, we identified interacting RNAs in HEK 293 cells transfected with 148 ADAR1 isoforms using RIP-seq. On the other hand, we identified RNAs that can become edited 149 by ADAR1 isoforms in MEF cells. To do so, ADAR1 deficient MEFs were transfected with 150 ADAR1 isoforms, and editing patterns were determined. Jointly, the two approaches give a 151 robust overview of the ADAR1-editome. 152

Further, we determined the impact of distinct motifs altering the localization and nucleic acid 153 binding properties of ADAR1-isoforms to assess their contribution to isoform-specific RNA- Next, MB-UV RIP-seq was performed for ADAR1p150 and ADAR1p110 isoforms in triplicates. 179

As a control, MOCK transfected cells were processed in parallel (in duplicates). 180

As expected, control IPs with untransfected samples showed no enrichment of ADAR1-181 specific substrates (Figure 1 -figure supplement 5 ). ADAR1p110-IP, when compared to 182

ADAR1p150-IP, displayed only mild enrichment of selected targets. This indicates that the 183 examined targets are prevalently bound and likely edited by ADAR1p150, a notion that was 184 confirmed for Azin1 (chr8:103841636) by Sanger sequencing (Figure 1 -figure supplement 5) . 185

We prepared triplicate libraries from ribo-minus selected RIP-samples. We obtained between 186 6 to 13 million uniquely mapped 75 bp paired-end reads ( Figure 1 -table 1 ). Next, peaks were 187 called for these mapped reads to identify ADAR1-isoform binding substrates and regions. 188

Peaks that also appeared in the empty-transfection control were subtracted from ADAR1p150 189 and ADAR1p110 peaks to exclude unspecific peaks. This resulted in 14,191 peaks for 190

ADAR1p150 and 2,921 peaks for ADAR1p110. Additionally, we excluded peaks shorter than 191 40 nt. Peak size is based on genomic rather than transcriptomic coordinates, thus ADAR1p150 192 peaks appear longer (the median size of ADAR1p150 peak is 2105 nt while ADAR1p110 peaks 193 are 524 nt on average) (Figure 1 -figure supplement 6 ). This is most likely caused by 194 cytoplasmically localized ADAR1p150 mostly binding to exonic regions. Consequently, these 195 peaks map across introns. We, therefore, calculated the median width of ADAR1p150 and 196 ADAR1p110 peaks and excluded all peaks longer than 3× the median peak size to omit 197 extremely long peaks. Using all the above-mentioned filtering criteria, we identified 10,255 198 peaks for ADAR1p150 and 2,289 peaks for ADAR1p110 ( Figure 1A ) (Figure1 -table 2). The 199 considerable discrepancy in the identified peak numbers can likely be explained by the higher 200 reaction dynamics of ADAR1p110 editing which occurs mostly co-transcriptionally in the 201 nucleus (Bentley, 2014; Licht et al., 2016) . Thus, detecting RNAs transiently bound by 202

ADAR1p110 might be less efficient than RNAs bound by cytoplasmic ADAR1p150 which likely 203 will be bound longer. The majority of ADAR1p110 peaks is located in introns which 204 corresponds well with its predominantly nuclear localization. In contrast, ADAR1p150 peaks 205 largely fall into exonic regions, including a large fraction of 3' UTRs ( Figure 1B) . A quite 206 prominent portion of intronic ADAR1p150 peaks is probably caused by the annotation of 207 peaks to genomic coordinates. Therefore, spanned introns also appear in the final peak 208 annotation, although they are not physically present in the peaks (Figure 1-figure supplement  209   7) . The ADAR1p150-and ADAR1p110 peak-set showed a limited overlap of 342 peaks for 210

ADAR1p110, 308 peaks for ADAR1p150, respectively (ADAR1p150 peaks overlap with more 211 than 1 ADAR1p110 peak) ( Figure 1C ). Thus about 12% of all ADAR1p110 peaks overlap with 212 ~2% of all ADAR1p150 peaks. This modest overlap is presumably caused by the limited binding 213 of ADAR1p150 to intronic regions compared to ADAR1-p110, which leads to a shifted 214 distribution of peaks for these two isoforms ( Figure 1D ). Almost 50% of ADAR1p150-peaks 215 span non-repetitive regions, whereas only 14% of ADAR1p110-peaks lie in non-repetitive 216 regions ( Figure 1E ). This observation greatly correlates with the gene-region distribution of 217 the peaks. Whilst ADAR1p110 binds mostly introns that contain a high portion of repeats, 218

ADAR1p150 also binds in 3'UTRs and exons. The majority of the interacting repeats accounts 219 for SINE elements in the case of both isoforms. However, ADAR1p150-peaks span also a 220 significant portion of simple repeats, LINE and DNA transposons ( Figure 1F ). In summary, our 221 RIP seq experiments identified distinct binding regions for ADAR1-isoforms. While 222

ADAR1p110 binds mostly intronic regions, ADAR1p150 also binds to exons and 3'UTRs. Restoring ADAR1 expression in A-to-I editing deficient cells.

After having shown a distinct binding preference for ADAR1p150 and ADAR1p110 isoforms in 239 human cells, we wanted to evaluate ADAR1 isoform-specific editing preferences. To do this 240 in an experimentally clean system, we turned to mouse cells that are deficient in ADAR1 and 241 ADAR2 and therefore showed no background editing (MEFs ADAR1-/-; ADAR2-/-). In these cells, we combining all samples and by considering all sites that were also covered by more than 5 256 reads in the negative control, we could identify 9.400 of more than 100.000 editing sites 257 known in mice and stored in REDIportal (Licht et al., 2019; Picardi et al., 2017) . Those sites 258 were subjected to further filtering to acquire reliable sets of editing sites for each ADAR1-259

isoform. First, we excluded all sites also detected in the RFP transfection control. Next, we 260 removed sites with low coverage (less than 5 reads per replicate) and editing rates <1%. 261

Finally, only sites edited in 2 out of 3 replicas and sufficiently covered in both isoform-datasets 262 (p150 and p110, average ≥ 10 reads) were subjected to further analysis ( ADAR1p150, and 25% of all sites are preferentially edited by ADAR1p110 ( Figure 2B) . ADAR1p150 is editing 3'UTRs with higher efficiency than ADAR1p110, consistent with its 282 predominant cytoplasmic localization. Preferences for intronic regions and 3' UTRs became 283 even more prominent when only sites with a clear preference for editing by one of the two 284 isoforms (sites that are edited 2,5 × better by one isoform than by another) were considered. 285

There, ADAR1p110 edited almost exclusively intronic sites whereas 3'UTRs became a 286 prominent target for ADAR1p150-editing ( Figure 2D ). 287

Next, we assessed editing in repetitive versus non-repetitive regions. ADAR1p150 edits more 288 nonrepetitive parts (≈ 23%) in comparison to ADAR1p110 (≈ 16%) ( Figure 2E ). 3'UTRs that are 289 preferentially edited by ADAR1p150 carry in general higher conservation than introns that are 290 mostly edited by ADAR1p110. The vast majority of editing in repetitive regions by both ADAR1 291 isoforms occurs in short interspersed nuclear (SINE) elements ( Figure 2F . Closer examination 292 of editing events in repeats revealed also editing in LTR and long interspersed nuclear (LINE) 293 elements ( Figure 2F ). Moreover, we studied ADAR1-editing in 'hyperedited' regions. Those 294 regions contain many potential editing sites and frequently bear repetitive regions, in 295 particular inverted SINEs. In our dataset, ADAR1p150 was a considerably greater editor of 296 those hyperedited regions, which was becoming more evident when only sites with a clear 297 preference for ADAR1p150 or ADAR1p110-editing (sites that are edited 2,5 × better by one 298 isoform than another) were considered ( Figure 2G ). Most mRNA molecules spend the 299 majority of their lifespan in the cytoplasm. Thus, ADAR1p150, which is mainly cytoplasmic, 300 might have longer access to hyperedited regions than nuclear ADAR1p110. 301

Thus, our data show that ADAR1p110 edits mainly intronic repetitive regions, while 302

ADAR1p150 is a powerful editor of 3'UTRs and hyperedited regions. ADAR1p150 is also more 303 involved in editing of non-repetitive and thus potentially more conserved editing sites. 304 305 306 307 However, nuclear ADAR1p150 did not edit any of the selected sites, whereas cytoplasmic 346

ADAR1p110 and even cytoplasmic ADAR2 did edit these substrates, as shown in the Sanger 347 sequencing chromatograms. Surprisingly, the P193A mutation of ZBDa did not affect editing 348 at both studied sites, which showed the same editing as wt ADAR1p150 ( Figure 3B ). This 349 finding suggested that the specificity of ADAR1p150 isoform is prevalently driven by its 350 cellular localization and, thus, might be substituted by any cytoplasmic editase. 2). Using these normalized values, we could validate the majority of ADAR1p150-selected 397 sites. However, individual sites in Tra2a, Rpa1 and Rad23 were preferentially edited by ADAR2 398 and cytoplasmic ADAR2. Our initial experiments did not include ADAR2, and hence, the 399 occurrence of sites ADAR2-editing sites next to ADAR1p150 sites cannot be excluded. 400

Nonetheless, the editing sites in Rpa1 and Rad23a showed higher editing rates for 401

ADAR1p150 than ADAR1p110 ( Figure 3C ). Quite surprisingly, the situation was less clear for 402

ADAR1p110-targets. Selected sites of Cdk13, Hnrnpf and Fbxl18 showed a clear preference 403

for ADAR1p150, while the sites in Gpalpp1 and Eif1ax showed a comparable preference for 404

ADAR1p150 and ADAR1p110. Most interestingly, many sites that could be well-edited by 405

ADAR1p110 were also highly edited by ADAR2, indicating preferential nuclear-editing of the 406 selected sites ( Figure 3C ). Employing the amplicon-seq approach, we confirmed that ADAR1p150 specificity is indeed 430 partially driven by its cytoplasmic localization. Based on this data, we suggest that mainly To determine a possible overlap between mouse substrates edited by ADAR1 isoforms and 443 human sequences bound by ADAR1 isoforms, we defined the overlap between both gene 444 sets. We found an overlap of 445 genes for ADAR1p150 and 153 genes for ADAR1p110. This 445 difference could be caused by the lower number of peaks detected by ADAR1p110 RIP-seq. 446

Editing is very abundant in human and occurs in nearly all genes due to the high repeat 447 content of the human transcriptome (Bazak et al., 2014a) . Thus, we next focused on genes 448 that were edited in non-repetitive regions. This resulted in 290 genes edited by ADAR1p150 449 and 142 genes edited by ADAR1p110. This way, we expected to find ADAR1 isoform-specific 450 editing targets that are conserved between mouse and human. In doing so, we could identify 451 26 overlapping genes edited or bound by ADAR1p110. Strikingly, more than half (151) In this study, we provide a robust overview of ADAR1-isoform-specific binding and editing 470 characteristics. We identify RNAs that are selectively bound and edited by either ADAR1 471 isoform, while a large fraction of RNAs can be bound and edited by either isoform. Most 472 interestingly, our study revealed that cellular localization is a major factor driving isoform 473 specific editing patterns. Still, editing of isoform-specific substrates is not exclusively 474 governed by the enzyme's localization, indicating that also other factors, most likely RNA-475 binding or competing RNA-binding proteins, can trigger ADAR1 substrate-specificity. 476 477 Using a modified RNA-IP (RIP)-seq protocol that involved methylene blue and UV cross-478 linking, we could improve the pull-down efficiency of ADAR-associated sequences. 479 ADAR1 is a dsRNA binding protein that mainly interacts with the RNA backbone (Ryter and 480

Schultz, 1998). Hence, conventional UV cross-linking is rather inefficient as this technique 481 creates covalent bonds between a protein and free unpaired RNA-bases that are limited in 482 dsRNA. We, therefore, applied methylene-blue mediated photo-crosslinking together with 483 UV-crosslinking to achieve robust binding between ADAR1 and its RNA-substrates. Methylene 484 blue intercalates into dsRNA, partially unwinds the dsRNA structure and thus, makes it 485 accessible for photo-crosslinking (Liu et al., 1996) . Moreover, additional UV-mediated cross-486 linking might further strengthen ADAR1-substrate binding and possible ssRNA-ADAR1 487

interactions (Wheeler et al., 2018) . Using this technique, we found distinct binding regions for 488 ADAR1-isoforms. While ADAR1p110 shows strongly enriched binding to introns, ADAR1p150 489 did efficiently capture exons and 3'UTRs. 490

We compared the outcome of our RIP-Seq results that were performed in HEK293 cells with 491 the currently published ADAR1-isoform-specific editome that was also generated in HEK293 sites edited by ADAR1p110. However, since the editing specificity at these sites was not 498 further confirmed we omitted them from further evaluation. 499

Next, using BEDTools-intersect (v2.30.0), we identified editing sites that directly overlap with 500 RIP-seq peaks identified by us (Quinlan and Hall, 2010) . Indeed, we found an overlap of ~25% 501 between sequence peaks defined by ADAR1p150 binding and editing sites that were 502 selectively edited by ADAR1p150 as detected in the Sun et al., study (Sun et al., 2021) (Figure  503 6A). We also found an overlap of ~19% between sequences bound by ADAR1p150 and editing 504 sites targeted by both ADAR1p150 and p110. Consistently, the overlap between ADAR1p110-505 peaks and both examined editing site sets was considerable smaller: ~10 % with ADAR1p150 506 editing sites and ~ 12,5 % with the combined ADAR1p150 & p110 editing site set ( Figure 6B ) 507 ( Figure 6 -Table 2) . Overall, the data show that ADAR1 p150 RIP-seq peaks have a larger 508 overlap with sites edited by p150 while there seems less overlap with sites precipitated by 509 ADAR1 p110 and those that are edited by ADAR1p150 and ADAR1p110. However, it should 510 also be noted that RIP-seq identified significantly fewer peaks for ADAR1p110. This 511 observation, together with preferential intronic binding of ADARp110 versus ADAR1p150 512 binding to exonic, 3'UTR regions, indicates that the identified peaks provide a valuable 513 dataset of high specificity. Still, we cannot exclude that overexpression of ADAR1 leads to 514 promiscuous binding and, hence ADAR1p110 could bind substrates that are otherwise 515 exclusive for ADAR1p150. However, apart from the N-terminus of ADAR1p150, ADAR1-516 isoforms are identical (George and Samuel, 1999; Patterson and Samuel, 1995) . Therefore, an 517 overlapping affinity for the same substrates can be expected. RIP-seq of endogenous ADAR1 518 might improve the substrate-specificity of the procedure. Nonetheless, since endogenous 519

ADAR1p150 is expressed at a very low level unless the cells are stimulated with IFN (Patterson 520 and Samuel, 1995) and specific antibodies recognizing only the N-terminus of ADAR1p150 are 521 scarce. Therefore, experimental conditions would require adequate optimization to 522 specifically detect binding sites of ADAR1 isoforms. 523 524 After having established ADAR1 isoform-specific binding substrates in human cells we also 525 examined ADAR1-isoform specificity in the mouse. Having the advantage of ADAR deficient 526 mouse cells as a tool, we restored ADAR1 expression in those cells. Here the ADAR1-isoform-527 specific editome was determined to gain insight into isoform-specific targets. 528 529 More than 100,000 known editing sites identified mostly in mouse brain are currently 530 available on REDIportal (http://srv00.recas.ba.infn.it/atlas/search.html) (Picardi et al., 2017) . 531

Of these, we detected 9,400 sites by restoring ADAR1-mediated editing in ADAR-deficient 532

MEFs. Using additional stringent filtering criteria, we obtained a comprehensive, high-quality 533 dataset of more than 3000 editing sites. We found ~2000 editing sites efficiently edited by 534 one or another ADAR1-isoform. While ADAR1p110 edits almost exclusively in introns, 535

ADAR1p150 seems a powerful editor of 3'UTRs. As expected, the majority of editing sites is 536 located in SINE elements. Nevertheless, ADAR1p150 showed a bigger portion of editing in 537 non-repetitive regions, which suggests that ADAR1p150-mediated editing might be more 538 conserved among species. Furthermore, ADAR1p150 is also more involved in the editing of 539 hyperedited regions, most probably due to the spatio-temporal advantage of its cytoplasmic 540

Although we identified ADAR1p110 editing sites, we cannot fully exclude that these sites 542 might also be edited by ADAR1p150. The RNA-seq depth is, in general, lower in intronic 543 regions that are preferentially edited by ADAR1p110 sites therefore, deeper sequencing 544 might reveal editing mediated also by ADAR1p150. Indeed, the editing sites in Gpalpp1 545 (chr14:76090115) that were identified to be exclusively edited by ADAR1p110 could 546 subsequently be shown to be edited by ADAR1p150 as well. However, this required amplicon-547 seq with several times higher coverage ( Figure 3C . Therefore, similar characteristics must be common for human and mouse ADAR1p150-558 mediated editing. In our current study, we employed two different approaches: I) ADAR1-RIP-559 seq performed in human cells and II) isoform-specific reconstitution of ADAR1 editing in 560 mouse MEF cells. To further compare these two approaches, we focused on genes with 561 editing sites in non-repetitive regions, assuming that such sites may exhibit more genomic 562 conservation. While only ~18 % of such genes overlap between the two experimental 563 strategies for ADAR1p110, more than half of such genes are common between ADAR1p150 564

ADAR1p150 substrates, including a highly-conserved editing site in Azin1 (chr15:38491612) 566 recently described as ADAR1p150 substrate also by Kim Still, it appears unlikely that one individual substrate is responsible for the phenotype of 571 limited editing sites that are exclusively edited by ADAR1p150 to prevent MDA5-activation 583 . In this study, mouse knockout models including an ADARp110 specific 584 knockout in the presence or absence of an additional ADAR2 deficiency were used to 585 characterize the remaining ADAR1p150-mediated editing in brain and thymus. We 586 intersected the exclusive-ADAR1p150 editing pattern identified in that study with the ADAR1 587 p150 editome identified here in MEFs. We found a big overlap between the datasets. 23 sites, 588 out of 53 in the brain and 312 out of 489 in the thymus ( Figure 6C ). Although all sites originally 589 identified in the brain showed a great preference for ADAR1p150-editing in MEFs, none of 590 them was exclusively edited by p150 in our setup ( Figure 6D ). In the thymus editing set, we 591 found 23 sites exclusive for ADAR1p150 editing and additional 22 sites with a high (more than that substitute for disrupted binding by mutated ZBDa. ADAR1p150 is interferon-inducible, 659 thus, once activated, ADAR1p150 is produced in high amounts which then leads to a great 660 amount of cytoplasmic editase that massively edits substrates in the cytoplasm. This might 661 explain the lack of phenotype in ADAR1 P195A/ P195A mice and the lack of AGS patients carrying 662 homozygous P193A mutations (Maurano et al., 2021b; Rice et al., 2017) . 663

In our study, we determined the characteristic of ADAR1-isoform-specific editing and binding. 664

We obtained a comprehensive, high-quality dataset that identifies specific and overlapping 665 substrates between ADAR1p150 and p110. Our data indicate that editing specificity of 666 ADAR1p110-mediated editing ratios detected in MEFs for sites that overlap with ADAR1p150 691 sites in the brain identified by . extracted with 1 ml of TRIFAST as described above for general RNA extraction. 738

For formaldehyde cross-linking, 0,1 % formaldehyde in 1 × PBS was added to cells and 740 incubated for 10 minutes at room temperature with gentle mixing. Cross-linking was 741 quenched by adding one-tenth of quenching buffer (2.5 M glycine and 25 mM Tris) (Ricci et 742 al., 2014) . 743

Methylene Blue cross-linking was performed by adding 18 ml of 3µg/ml Methylene Blue in 1 744 × PBS. Cells were kept on ice and subsequently exposed to visible light for 30 minutes (Kaiser 745

Prolite 5000). Ultraviolet cross-link 2 × 800 mJ was additionally applied (UVP Crosslinker, 746

Analytic Jena GmbH, Jena, Germany). 747

Cross-linked cell pellets were kept at -80 °C until cell lysis. The IP-protocol was modified from 748

Ricci et al. (Ricci et al., 2014) . In short, cell pellets were lysed in 1.2 ml (per 2×150 mm dishes) 749 of hypotonic lysis buffer (20 mM Tris-HCl, pH 7.5, 15 mM NaCl, 10 mM EDTA, 0.5% NP-40, Reads were adapter-clipped using Cutadapt (Martin, 2011) and aligned to the mouse genome 849 mm10 with STAR using public server at usegalaxy.org ( Domain organisation of active mammalian ADAR proteins: ADARs contain a conserved Cterminal catalytic-deaminase domain and two or three double-stranded RNA binding domains (dsRBD). Whilst both ADAR1 isoforms have Z-DNA binding domain b (Zb), full-length ADAR1p150 isoform also bears N-terminal Z-DNA binding domain a (Za) and a nuclear export signal (NES). All ADAR variants possess nuclear localisation signals (NLS). While the NLS of ADAR2 is localised at the N-terminus, the NLS of ADAR1 is bimodular assembled around the third dsRBD. RIP-seq experimental workflow: A) HEK 293 cells were transfected with ADAR1p150-, ADAR1p110 -plasmid DNA or mock-empty transfection. The next day cells were subjected to selected cross-linking and harvested. The input sample was collected, and the remaining cell suspension was submitted to flag-IP. RNA was extracted from input-and IP-samples, and enrichment of selected editing substrates in IP fractions was examined using qPCR. Ribosomal RNA was depleted and RNA was converted to cDNA and subjected to high-throughput sequencing. The obtained reads were analysed via peak-calling. ADAR1p150 and ADAR1p110 RIP-seq were performed in triplicates and empty-mock transfection in duplicates. B) HEK 293 cells after exposure to methylene blue + UV cross-linking. 

A) qPCR evaluation of enrichment of ADAR1-targets in IP-fractions. The enrichment was normalised to the relevant input sample. RIP p110 and RIP p150 were conducted in triplicates and RIP MOCK in duplicate. B) Sanger sequencing traces of a selected editing site. Azin1 (chr8:103841636) is primarily edited by ADAR1p150. The reverse strand is sequenced. Consequently, an A to I event is seen as a T to C conversion in the chromatogram-indicated with an asterisk (*) Peaks can span several exons. A) The peak 13012 spans several exons of Azin1. B) Readcoverage of IP samples in the peak-spanning region. Peak 13012 is clearly specific for ADAR1p150. The peak is physically formed almost exclusively from exons; therefore, the actual size of the peak is dramatically smaller than the size calculated based on the genomic coordinates. Mouse embryonic fibroblasts (MEFs) generated from ADAR1 -/-, ADAR2 -/knockout mouse embryos, were electroporated with ADAR1p150, ADAR1p110 or RFP mammalian expression vector. 24 hours later cells were harvested and RNA was extracted. rRNA-depleted RNA-seq libraries were prepared and subjected to NGS.

Editing was examined at know-editing sites. Over-expressed ADAR1-isoforms showed typical cellular localisation and distinct editing patterns. A) ADAR1p150 is mainly localised to the cytoplasm, whereas ADAR1p110 is primarily nuclear. TRITC channel shows transfected constructs in confocal sections, and nuclear DNA is stained with DAPI (Scale bar: 20 µm). B) Editing site in the 3'UTR of Pum2 (chr12: 8750269) is edited by ADAR1p150 but not by ADAR1p110. Restoring ADAR1 expression in editing-deficient cells produces an authentic set of editing sites. A) Editing in known-editing sites detected in each sample. Only few A to G transitions were detected in the negative control: 78 sites. Editing rises dramatically upon transfection with any ADAR1-isoform, ranging from 1121 to 5395 detected sites. Sites covered by ≥ 5 reads in at least two replicas showing ≥ 1% editing. B) rRNA reads in each sample. The first replica of ADAR1p150 shows a remarkably high rRNA fraction (77 %). While the negative control also has a prominent rRNA fraction the remaining samples exhibit a minimal rRNA portion (less than 1 %). C) Filtering strategy for the master set generation. All detected editing sites sufficiently covered (≥ 5 reads) in the negative control were combined, resulting in 9.400 sites. Next, sites found edited in the negative control were subtracted. Sites covered by at least 5 reads and showing an editing ratio of at least 1% in 2 out of 3 replicas were collected. Only sites with sufficient coverage in datasets of both isoforms (≥ 10 reads in average) formed the final 'master' dataset containing 3.073 editing sites. D) Donut plot showing a portion of editing sites passing the filtering criteria. Heat maps of normalized editing ratio of all editing sites identified by amplicon-seq that are edited ≥ 0,5%. Apart from selected sites, amplicon-seq covers several closely located editing sites. Surprisingly, some very close sites show dramatically distinct editing patterns. A) Editing targets selected as preferentially edited by ADAR1p150 based on the ADAR1-editome obtained in MEFs; B) Editing targets selected as preferentially edited by ADAR1p110 based on the ADAR1-editome in MEFs.

p150 p110 nuc-p150 cyt-p110 mZ-p150 Ad2 cyt-Ad2 

The Galaxy platform for accessible, reproducible 890 and collaborative biomedical analyses

Breaching Self-Tolerance to

A bimodular 896 nuclear localization signal assembled via an extended double-stranded RNA-binding domain 897 acts as an RNA-sensing signal for transportin 1

A-to-I RNA editing occurs at over a hundred 901 million genomic sites, located in a majority of human genes

Genome-wide analysis of Alu editability

Accumulation 906 of nuclear ADAR2 regulates adenosine-to-inosine RNA editing during neuronal 907 development

Coupling mRNA processing with transcription in time and space

The 911 zalpha domain of the editing enzyme dsRNA adenosine deaminase binds left-handed Z-RNA 912 as well as Z-DNA

Neuropilin-1 facilitates SARS-CoV-2 916 cell entry and infectivity

fastp: an ultra-fast all-in-one FASTQ 918 preprocessor

Characterization of human 921 disease phenotypes associated with mutations in TREX1

ADAR1 interaction with Z-925

RNA promotes editing of endogenous double-stranded RNA and prevents MDA5-dependent 926 immune activation

Dynamic association of RNA-editing enzymes with the nucleolus

STAR: ultrafast universal RNA-seq aligner

Modulation of ADAR1 980 editing activity by Z-RNA in vitro

The Sequence Alignment/Map 983 format and SAMtools

A 985 high resolution A-to-I editing map in the mouse identifies editing events controlled by pre-986 mRNA splicing

Adenosine to Inosine editing 988 frequency controlled by splicing efficiency

RNA editing by ADAR1 prevents MDA5 sensing of 992 endogenous dsRNA as nonself

Beyond CLIP: advances and opportunities to measure RBP-994

Detection of double-stranded 996

RNA-protein interactions by methylene blue-mediated photo-crosslinking

A type I interferon signature 999 identifies bilateral striatal necrosis due to mutations in ADAR1

Investigating RNA editing in 1002 deep transcriptome datasets with REDItools and REDIportal

ViennaRNA Package 2.0

The RNA-editing enzyme ADAR1 1009 controls innate immune responses to RNA

Cutadapt removes adapter sequences from high-throughput sequencing 1012 reads

Protein kinase R and the integrated stress response drive immunopathology caused 1015 by mutations in the RNA deaminase ADAR1

Protein kinase R and the integrated stress response drive immunopathology caused 1019 by mutations in the RNA deaminase ADAR1

Mutations in the adenosine deaminase ADAR1 that prevent 1022 endogenous Z-RNA binding induce Aicardi

Functions and regulation of RNA editing by ADAR deaminases

A-to-I editing of coding and non-coding RNAs by ADARs

Enhanced AZIN1 RNA editing and 1030 overexpression of its regulatory enzyme ADAR1 are important prognostic biomarkers in 1031 gastric cancer

Expression and regulation by interferon of a double-1033 stranded-RNA-specific adenosine deaminase from human cells: evidence for two forms of 1034 the deaminase

Comprehensive analysis of RNA-Seq data reveals extensive RNA editing in a 1037 human transcriptome

Isoforms of RNA-Editing Enzyme ADAR1 Independently Control Nucleic Acid Sensor

REDIportal: a comprehensive 1043 database of A-to-I RNA editing events in humans

REDItools: high-throughput RNA editing detection made 1046 easy

A left-1048 handed RNA double helix bound by the Z alpha domain of the RNA-editing enzyme ADAR1

CRM1 mediates 1051 the export of ADAR1 through a nuclear export signal within the Z-DNA binding domain

BEDTools: a flexible suite of utilities for comparing 1054 genomic features

Staufen1 senses overall transcript secondary structure to 1057 regulate translation

Mutations in ADAR1 1060 cause Aicardi-Goutieres syndrome associated with a type I interferon signature

Genetic, Phenotypic, and Interferon

Biomarker Status in ADAR1-Related Neurological Disease

Molecular basis of double-stranded RNA-protein 1067 interactions: structure of a dsRNA-binding domain complexed with dsRNA

Structure-function 1070 analysis of the Z-DNA-binding domain Zalpha of dsRNA adenosine deaminase type I reveals 1071 similarity to the (alpha + beta) family of helix-turn-helix proteins

AZIN1 RNA editing confers cancer 1077 stemness and enhances oncogenic potential in colorectal cancer

Neuropilin-1 is 1080 expressed by endothelial and tumor cells as an isoform-specific receptor for vascular 1081 endothelial growth factor

Nucleocytoplasmic distribution of 1083 human RNA-editing enzyme ADAR1 is modulated by double-stranded RNA-binding domains, 1084 a leucine-rich export signal, and a putative dimerization domain

Decoupling expression and editing 1088 preferences of ADAR1 p150 and p110 isoforms

Adenosine-to-inosine editing of endogenous Z-1092 form RNA by the deaminase ADAR1 prevents spontaneous MAVS-dependent type I 1093 interferon responses

Comprehensive RNA editome reveals that edited Azin1 partners with DDX1 to 1096 enable hematopoietic stem cell differentiation

Stress-induced apoptosis associated with null mutation of ADAR1 RNA editing 1099 deaminase gene

Advances and challenges in the 1101 detection of transcriptome-wide protein-RNA interactions

RIP-Seq data analysis to determine RNA-protein 1104 associations

CrossMap: a 1106 versatile tool for coordinate conversion between genome assemblies

1109 (2010). p53 loss promotes acute myeloid leukemia by enabling aberrant self-renewal

ADAR1p150 and ADAR1p110-mediated editing ratios detected in MEFs for sites found edited by ADAR1p150 in the thymus

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

µl/sample Turbo DNaseI for 10 min at 37 °C (Invitrogen TM , Thermo Fisher Scientific, Waltham, 756 Massachusetts). NaCl was adjusted to 150 mM, and the lysate was cleared by two 757 centrifugation steps at 15.000 g for 10 minutes at 4°C. The input samples were collected. The 758 lysate was incubated for 2 hours at 4°C with 180 µl of anti-flag agarose beads (50 % slurry, 759 ANTI-FLAG® M2 Affinity Gel, Merck, Kenilworth, NJ, USA), prewashed 3 × with 1 ml isotonic 760 wash buffer (IsoWB) (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.1% NP-40). Next, beads 761 were washed 2 × with 1 ml IsoWB + 0.1% SDS and 0.1% sodium deoxycholate, then 2 × with 1 762 ml with IsoWB. Protein-RNA complexes were eluted with 300 µl of PK-7M Urea buffer ( Workflow of amplicon-seq experiment: Total RNA was transcribed into cDNA. Amplicons were prepared via target-specific PCR with PCR bearing partial Illumina adaptor sequence. Next, amplicons were further amplified and barcoded by a second PCR reaction with primers containing complementary adaptor sequence and unique index-barcodes. Amplicons were sequenced, and the obtained reads were examined for editing events at known-editing sites. Heat maps of all editing sites detected by amplicon-seq in selected ADAR1p150 targets and folding prediction of substrates. Structure predictions were generated using Vienna RNA Package. Magnifications of edited regions display detected editing sites in the structure (labelled with the last 3 digits of the genomic coordinate).