Distinct interactomes of ADAR1 nuclear and cytoplasmic protein isoforms and their responses to interferon induction Authors: Dragana Vukić, Anna Cherian, Salla Keskitalo, Yih Tyng Bong, Martin Marônek, Leena Yadav, Liam P Keegan, Markku Varjosalo, Mary A O’Connell Published: 2024 ABSTRACT AbstractThe RNA editing enzyme adenosine deaminase acting on RNA 1 (ADAR1) is essential for correct functioning of innate immune responses. The ADAR1p110 isoform is mainly nuclear and ADAR1p150, which is interferon (IFN) inducible, is predominately cytoplasmic. Using three different methods – co-immunoprecipitation (co-IP) of endogenous ADAR1, Strep-tag co-IP and BioID with individual ADAR1 isoforms – a comprehensive interactome was generated during both homeostasis and the IFN response. Both known and novel interactors as well as editing regulators were identified. Nuclear proteins were detected as stable interactors with both ADAR1 isoforms. In contrast, BioID identified distinct protein networks for each ADAR1 isoform, with nuclear components observed with ADAR1p110 and components of cytoplasmic cellular condensates with ADAR1p150. RNase A digestion distinguished between distal and proximal interactors, as did a double-stranded RNA (dsRNA)-binding mutant of ADAR1 which demonstrated the importance of dsRNA binding for ADAR1 interactions. IFN treatment did not affect the core ADAR1 interactomes but resulted in novel interactions, the majority of which are proximal interactions retained after RNase A treatment. Short treatment with high molecular weight poly(I:C) during the IFN response resulted in dsRNA-binding-dependent changes in the proximal protein network of ADAR1p110 and association of the ADAR1p150 proximal protein network with some components of antiviral stress granules. Introduction Adenosine deaminase acting on RNA 1 (ADAR1) is an RNA editing enzyme that hydrolytically deaminates adenosine to inosine in double-stranded RNA (dsRNA) [for review (1)]. Its catalytic activity is crucial for distinguishing between ’self’ and ‘non-self’ RNA, thus playing a critical role in innate immunity (2). ADAR1 exists in two isoforms; the constitutively expressed p110 isoform, mainly localized in the nucleolus (3), and the interferon (IFN)-inducible p150 isoform, which is predominantly cytoplasmic. Nuclear localization is determined by the nuclear localization signal located in the third dsRNA binding domain (dsRBD3) (4). The cytoplasmic presence of ADAR1 is established by the nuclear export signal in the Z domain at the amino terminus that is present only in ADAR1p150 (5). During oxidative stress, ADAR1p150 localizes to stress granules (SGs) in the cytoplasm via its Z domain (6). The cytoplasmic ADAR1p150 isoform is required to prevent aberrant innate immune responses [for review (7)]. The role of ADAR1 in innate immunity was elucidated with mouse genetic models. The Adar null mutant is embryonically lethal and dies by day E12.5 with hemopoietic defects liver disintegration, and a strong, aberrant IFN response (8,9). This lethality is rescued to live birth in Adar Mavs double mutants also lacking the Mavs (mitochondrial antiviral signaling) protein required for signaling from dsRNA-activated retinoic acid-inducible gene I (Rig-I)-like receptors (RLRs) (2,10,11). In another mouse model, a mutation in Adar that generates a catalytically inactive protein was combined with a mutation in the gene encoding melanoma differentiation-associated protein 5 (Mda5, gene name Ifih1), which is the RLR activated by long dsRNA (12). This double mutant completely rescues the Adar phenotype, is fertile, and has a normal lifespan. The differences between the Adar null and catalytically-inactive mutants demonstrate that the Adar1 protein itself also has biological functions independent of its RNA editing activity. Recently it was shown that a triple mutant combining an allele of Adar mutated to generate only the p110 isoform with mutations in Ifih1 and in the gene encoding Pkr (Eif2ka2), rescues the Adar phenotype independent of the editing activity (13,14). While the phenotypes of Adar mutant mice are consistent for each particular allele, the effect of ADAR1 depletion in cultured cells varies. In some cell lines, ADAR deletion results in cell death, whereas in others, cells survive without it (15); cell death is rescued by depleting MDA5 or, in some cell types, by additional depletion of PKR. Given the importance of the ADAR1 editing-independent function and the variations in ADAR1’s interactions and effects across different cells, we decided to explore ADAR1 protein interactions with three different enrichment methods combined with liquid chromatography-tandem mass spectrometry (LC–MS/MS). This was performed in human cell lines so we could compare our results which those previously published on the ADAR1 interactome (16,17). First, we mapped the interactome of endogenous ADAR1 in human A549 cells previously used to study ADAR1 (15), in the presence and absence of RNase A treatment. To discern isoform-specific interactions, we created stable cell lines with the Flp-In™ T-REx™ 293 cell line, expressing either ADAR1p110, ADAR1p150 or mutant ADAR1 isoforms incapable of dsRNA binding (EAA). We utilized the Strep II-HA affinity tag to study stable interactions of individual ADAR1 isoforms, with both data-dependent acquisition (DDA) and data-independent acquisition (DIA) modes of MS/MS. To capture spatial and temporal interactions, we employed BioID, tagging ADAR1 with biotin ligase BirA* which labels proteins within an approximate radius of 10 nm. Lastly, since ADAR1p150 expression is normally induced by IFN (18) we also examined the ADAR1 isoform interactomes under IFN stimulation with IFNα or HMW poly (I:C) (henceforth referred to as p(I:C)) treatments. With these different methods (19), we have identified both known and novel ADAR1 interactors. ADAR1 engages in a broad range of interactions, aligning with its known biological roles. IFN stimulation expands the ADAR1 protein interactome to include viral replication inhibitors and L1 regulators. Analysis of the proximal protein network reveals that after IFN and p(I:C) treatments, the ADAR1 interactome changes: ADAR1p110 interacts with nuclear export machinery and dsRNA binding proteins (dsRBPs), while ADAR1p150 associates with antiviral proteins and with cytoplasmic condensates. Mutations affecting dsRNA binding alter interactions differently depending on the ADAR1 isoforms, notably impacting ADAR1p110’s nucleolar interactions. Materials and methods Cells, culture conditions and treatments Human Flp-In™ T-REx™ 293 cell lines (Invitrogen) were maintained in Dulbecco’s modified Eagle’s medium DMEM) Low (# LM-D1102/500, Biosera), supplemented with 10% tetracycline-free fetal bovine serum (FBS) (#P30-3602, Biosera). A549 cells were maintained in DMEM (# LM-D1110/500, Biosera) supplemented with 10% FBS (#FB-1001B/500, Biosera) and 1% non-essential amino acids (#M7145-100M, Sigma–Aldrich). To each culture media 1% penicillin–streptomycin solution (#XC-A4122/100, Biosera) was added. For induction of the IFN response, cells were treated with 2000 IU/ml of human IFNα 2a (#SRP4594, Merck) or 0.3 or 0.5 μg/ml of poly(I:C) (HMW) / LyoVec™ (#tlrl-piclv, InvivoGen). For the RNase A experiments, cells were treated with 2000 IU/ml of Human IFN-alpha 2 (alpha2a) recombinant protein (# 300–02AA-100UG, Thermo Fisher Scientific). The specific time points and concentration used for each experiment are described in the figure legends and main text. Preparation of expression constructs and stable cells Coding sequences for ADAR1p150 and ADAR1p110, both wild-type (WT) and EAA mutant, were amplified with Phusion™ High-Fidelity DNA Polymerase (#F530L, Thermo Scientific), following the manufacturer’s instructions with gene-specific primers (Supplementary Table S5) from Generi Biotech. Gateway™ recombination technology (Thermo Fisher) was used to generate the constructs. The vectors were pTO-Strep II-H (20) and pTO-BirA*-MYC and the ADAR1 inserts encoded N-terminal epitope tags. The ADAR1p110 EAA mutant was a kind gift from Kazuko Nishikura (21). Flp-In™ T-REx™ 293 cell line (Invitrogen) was co-transfected with 300 ng of the ADAR plasmid (pTO-Strep II-H or pTO-BirA*-MYC) and 2.7 μg of pOG44 vector for Flp-recombinase expression (Invitrogen) with TurboFect™ Transfection Reagent (#R0532, Thermo Scientific) or Lipofectamine™ 3000 Transfection Reagent (#L3000015, Invitrogen) as per the manufacturer’s instructions. After 24 h, cells were transferred to a 150 mm cell culture dish and selected for 2–3 weeks with 100 μg/ml of hygromycin B Gold (#ant-hg-1, Invivogen). Individual clones were screened by induction with 1 μg/ml of doxycycline for 24 h followed by immunoblotting. Positive clones were further expanded and used. Generation of the Flp-In™ T-REx™ 293 cell line expressing the green fluorescent protein (GFP) control (BirA*-MYC or Strep II-HA) was previously described (22). RNA extraction, cDNA synthesis, and quantitative PCR Total RNA was isolated with TriPure reagent (#T9424, Merck) following the manufacturer’s instructions and treated with 2 U of TURBO™ DNase (#AM2239, Invitrogen) for 20–30 min at 37°C in the presence of RiboLock RNase Inhibitor (#EO0382, Thermo Scientific). DNase-treated total RNA was purified with phenol-chloroform extraction and an ethanol precipitation with sodium acetate. Precipitated RNA (0.5 μg) was then reverse transcribed with RevertAid Reverse Transcriptase kit (#EP0442, Thermo Scientific) and random hexamers, following the protocol for first strand cDNA synthesis on the manufacturer’s website. Quantitative-PCR was performed on a LightCycler® 480 Instrument II (Roche) with LightCycler® 480 SYBR Green I Master (#04 887 352 001, Roche) with the following cycling conditions: 95°C for 5 min; followed by 40 cycles of 95°C for 10 s; 56°C for 20 s; 72°C for 10 s; and 40°C for 10 s. Each reaction contained 10 ng of cDNA and gene-specific primers (Supplementary Table S5). Samples were analyzed with three biological and three technical replicates and transcripts were normalized to two housekeeping transcripts (GAPDH and RPL13). For relative quantification of gene expression the comparative CT method (ΔΔCT) was used (23). Statistical analysis was performed with a two-sided, unpaired t-test in the R programming language (24,25) for mean comparisons between IFNα and non-treated (NT) samples. Immunofluorescence Cells were seeded on coverslips coated with 0.1% gelatin in 24-well plates. Following treatment, the cells were fixed with 4% paraformaldehyde for 20 min at room temperature (RT) and permeabilized with 0.2% Triton X-100 for 20 min. Blocking was performed with 5% bovine serum albumin for 1 h at 37°C. The coverslips were then incubated with primary antibodies (Supplementary Table S6) in a humid chamber at 4°C overnight. Afterwards, the coverslips were washed thrice with phosphate-buffered saline (PBS) and incubated with secondary antibodies (Alexa Fluor 488 goat anti-mouse IgG (H + L): Thermo Fisher Scientific, cat. A-11001, 1:500, Alexa Fluor 594 goat anti-rabbit IgG (H + L): Thermo Fisher Scientific, cat. A-11012, 1:500) for 1 h at 37°C. Subsequently, they were washed three times with PBS, once with distilled water, and mounted on slides (Superfrost Plus, Thermo Fisher Scientific) using ProLong Gold Antifade reagent with DAPI (Thermo Fisher Scientific). The slides were left to solidify overnight in the dark. The samples were visualized with a Zeiss LSM 780 confocal microscope with a 63× (1.4 numerical aperture) oil-immersion objective lens. The raw images were processed in Fiji (26). Immunoblotting of total cell lysate Samples were lysed for 30 min in lysis buffer [0.5% NP40; 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonicacid (HEPES), pH 8.0; 5 mM ethylenediaminetetraacetic acid (EDTA); 150 mM NaCl] supplemented with 1× Halt™ Protease Inhibitor Cocktail, EDTA-Free (#78 439, Thermo Fisher) and 1× Halt™ Phosphatase Inhibitor Cocktail (#78 427, Thermo Fisher) unless specified otherwise. Lysis was followed by centrifugation at 18 000 g for 20 min and total protein quantification was performed with the Bicinchoninic acid assay (BCA) Protein Assay Kit (#23 225, Pierce™). Samples were denatured at 95°C for 5 min in Laemmli Sample Buffer and loaded on Tris-HCl sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) or a mPAGE® 4–12% Bis-Tris Precast Gel (#MP41G10, Merk). Separated proteins were transferred to a nitrocellulose membrane (#10 600 003, VWR) with a Mini Trans-Blot® Cell (Bio-Rad) for 1 h at 100 V. Membranes were blocked and subsequently incubated with primary antibodies overnight at 4°C, and with secondary antibodies for 1 h at RT. Antibody information is provided in Supplementary Table S6. Clarity Western ECL Substrate (#1 705 061, Bio-Rad) and UVITEC ChemiDoc imaging device were used for detection. Endogenous co-immunoprecipitation and RNase A treatment Following 30 h of treatment with 2000 IU/ml of IFNα2a or 0.5 μg/ml of poly(I:C), A549 cells were processed immediately or snap-frozen in liquid nitrogen and stored at −80°C. Lysis was performed for 30 min in lysis buffer (50 mM Tris–HCl, pH 8.0; 150 mM NaCl; 0.5% NP40 and 1 mM EDTA, pH 8.0, supplemented with protease cOmplete™ Protease Inhibitor Cocktail (#11 697 498 001, Sigma). For the RNase A experiment, EDTA was omitted from the lysis buffer. Equal amounts of proteins, quantified with the BCA Protein Assay Kit (#23 225, Pierce™), were used for subsequent steps of IPs and part was set aside for ‘input’ sample. For the RNase A digestion experiment, 30 μg/ml of RNase A (Thermo, EN0531) was added to 3 mg of total protein and incubated for 1 h at 4°C. Samples without the RNase A digestion, were incubated with 30 U/ml of RiboLock RNase Inhibitor (Thermo Scientific™, EO0381). Samples for the endo-IP experiment I and II were pre-cleared with DynabeadsTM Protein G (#10003D, Thermo Fisher) for 1 h and subsequently incubated for an additional hour with ADAR1 (#ABIN2855100, Antibodies.com) or IgG control (#30000–0-AP, Proteintech) antibodies. Equilibrated beads (DynabeadsTM Protein G) were added to the lysate-antibody solution and incubated overnight on a rotation shaker at 4°C. For the RNase A endo-IP experiment, digested or undigested sample lysate was incubated with ADAR1 or the IgG control antibody overnight and equilibrated DynabeadsTM Protein G were added next day for 1 h. For the RNase A experiment, part of the sample was taken for RNA extraction before adding beads. Following the indicated incubations, the beads were washed 6× with the lysis buffer prepared for LC–MS/MS analysis. For the RNase A experiments beads were washed 3× with lysis buffer and 3× with wash buffer (50 mM Tris–HCl, pH 8.0; 200 mM NaCl; and 1 mM EDTA, pH 8.0). Part of beads (∼10%) were retained and analyzed by SDS-PAGE electrophoresis and immunoblotting as described above. RNase A validation experiments were performed as described above, with slight modifications; 2.8 mg/ml of total protein was incubated with RNase A (10 μg/ml) or RiboLock (20 U/ml). Following 1 h of incubation, RiboLock was added to the RNase A sample and endo-IP steps were as described above. Strep II-tag co-immunoprecipitation Flp-In™ T-REx™ 293 cell line with stably integrated Strep II-HA-N-tagged ADAR1 and GFP, were induced with doxycycline (1 μg/ml for 28 h) and subsequently treated with IFNα (2000 IU/ml for 24 h). Cells were lysed in 1 ml of lysis buffer [50 mM HEPES, pH 8.0; 5 mM EDTA; 150 mM NaCl; 0.5% NP40; 1 mM dithiothreitol (DTT) and 2× Halt™ Protease Inhibitor Cocktail, EDTA-Free (#78 439, Thermo Fisher)] for 40 min. After centrifuged at 12 000 rpm for 20 min, part of the supernatant was retained as an ‘input’ sample and remainder was incubated with 55 μl of equilibrated MagStrep ‘type3’ XT beads (#2–4090-002, IBA GmbH), overnight on a rotation shaker at 4°C. The next day, after six washes with the lysis buffer, part of the beads (10%) was used for immunoblot analysis (as described above) and the remaining volume was prepared for LC–MS/MS analysis. Proximity-dependent biotin identification labeling and purification Flp-In™ T-REx™ 293 cell line stably expressing pTO-BirA*-MYC-N-tagged ADAR1 or GFP were grown in 100 mm cell culture dishes (two per each condition). For induction, 1 μg/ml of doxycycline was added for a total of 28 h; 4 h after doxycycline induction, 50 μM of biotin (#B4501, Sigma) and IFNα2a (2000 IU/ml) were added to the cells. The next day, 4 h before cell collection, cells were treated with 0.5 μg/ml HMW poly (I:C)/LyoVec™ (#tlrl-piclv, InvivoGen). The cells were pelleted and three washes with PBS was performed and if not processed immediately after pelleting, the samples were flash-frozen and stored at −80°C. Affinity purification of biotinylated protein was performed as described in (27) with some minor modifications as described (28). The sonication step was performed with the same setting and time as described but with a 5/64 probe and QSonica, Q55A instrument. The cleared lysate was incubated overnight with 85 μl of magnetic streptavidin beads (#65 001, Thermo Fisher) on a rotating wheel at 4°C. Part of the beads was retained for immunoblotting analysis with HRP-conjugated streptavidin (Supplementary Table S6) as described (27) and the remainder prepared for LC–MS/MS analysis. Sample preparation and LC–MS/MS analysis Full details regarding the analyses and data evaluation are in Supplemental Materials. In summary, for BioID and RNase A endo-IP analysis, the bead-bound protein complexes were treated with AMBIC and trypsin, incubated at 37°C for 2 h, followed by reduction with Tris(2-carboxyethyl)phosphinehydrochloride (TCEP) and alkylation with Iodoacetamide (IAA). After an additional 16 h of trypsin incubation at 37°C, peptides were quenched with TFA, desalted with C18 MicroSpin columns and reconstituted in [0.1% (v/v) trifluoroaceticacid (TFA) and 1% (v/v) acetonitrile (#83640.320, VWR)] in high pressure liquid chromatography (HPLC) water. Samples were further diluted with HPLC water containing formic acid and loaded onto an Evotips. They were analyzed with the Evosep One liquid chromatography system coupled to a hybrid trapped ion mobility quadrupole Time-of-flight (TOF) mass spectrometer (Bruker timsTOF Pro, Bruker Daltonics) (29) via a CaptiveSpray nano-electrospray ion source (Bruker Daltonics). For Strep II-tag IP analysis, the bead-bound protein complexes were mixed with 2% SDS solution and heated to 50°C for 20 min after which the proteins were reduced with DTT at 95°C for 15 min. The resulting protein solutions were subjected to filter-aided sample preparation as previously described (30). The resulting peptides were analyzed by LC–MS/MS performed with the UltiMate 3000 RSLCnano system (Thermo Fisher Scientific) on-line coupled with Orbitrap Fusion Lumos spectrometer (Thermo Fisher Scientific). High-confidence interactors and label-free quantification High-confidence interactors (HCIs) for endo-IP and BioID experiments were defined as previously described (31). In summary, cut-off criteria for HCIs with SAINT (32) included Bayesian false discovery rate (BFDR) <0.05, Average spectral count ≥4, Fold Change >4, SAINT score >0.7 and <20% for CRAPome (33). If a protein presence in CRAPome was >20%, the average spectral count fold change limit was set to 3 or manually checked when spectral counts for bait (ADAR1) differed between conditions. For the Strep II-tag IP experiment, differential expression with the LIMMA statistical test was used, with the following criteria: log2FC relative to GFP >2; adj. P-value <0.01; average number of peptides ≥4 present in at least two biological replicates. Strep II-tag DDA dataset was further filtered against the CRAPome, proteins not passing criteria described above were removed from both DDA and DIA dataset. Ribosomal and cytoskeletal proteins that passed the criteria described above were removed manually before further analyses. For data pre-processing an in-house python and R-scripts were used. Raw intensity values were log2 transformed and median-normalized. Afterward, missing values were imputed with QRILC imputation (34). For sample group comparison, log2FC were obtained by the ratio of the bait-normalized means and the P-values were calculated with Student’s t-test. These values were used for generating the volcano plots with an in-house R-script. For ADAR1-normalized spectral counts, the average spectral counts for each prey was normalized to the average ADAR1 spectral counts for comparisons. The Gene Ontology (GO)-terms analysis, IFN-related proteins and protein network visualization GO-terms analysis was performed with The Database for Annotation, Visualization and Integrated Discovery (DAVID, version: v2024q2) (35). The background was set to ‘Homo sapiens’ and the ‘annotation chart’ option was chosen. Terms were filtered based on Bonferroni, Benjamini and false discovery rate (FDR) values, all set to be <0.01. A full list of the enriched terms and relevant statistical values for the BioID experiment are available in the Supplementary Table S2. IFN-related proteins were filtered with the Interferome database [version 2.01 (36)]. HCIs were searched for the term ‘Homo sapiens’ and the fold change was set to >2. Protein networks were constructed with STRING database version: 12.0 (37,38), both physical interactions and functional interactions with 0.4 confidence. Networks were visualized with Cytoscape [version 3.9.1 (39)]. Statistical analyses, visualization of the data All graphical data visualization in this manuscript was generated with RStudio (v. 2022.07.2 + 576 with R v. 4.2.2) and the ggplot2 R package (40). Graphical abstract and schemes displayed in Figures 1A, 2A, 3A and 4A, Supplementary Figure S1A were created with BioRender. Interactomes of ADAR1 isoforms during homeostasis. (A) Scheme depicting the experimental approach used for the identification of stable HCIs of ADAR1. Endogenous ADAR1 and associated proteins were immunoprecipitated (IP) with an ADAR1-specific antibody and subsequently analyzed by LC–MS/MS in the DDA mode. Doxycycline-inducible stable cell lines 293 T-REx expressing either ADAR1p110 (yellow) or ADAR1p150 (green) were generated, enabling isoform-specific Strep-tag IP. The bound proteins were subjected to LC–MS/MS in both DDA and DIA modes. This image was created with BioRender. (B) Venn diagram illustrating the overlaps between the HCIs of endogenous ADAR1 and Strep-tagged ADAR1 isoforms, with both DDA and DIA modes. (C) Functional protein network of HCIs identified for both endogenous ADAR1 and Strep-tagged ADAR1 isoforms [areas highlighted in color in panel (B)]. The network was generated in STRING with medium confidence and visualized with Cytoscape. Node colors represent isoform-specific interactors (grey– both ADAR1 isoforms, yellow – ADAR1p110 unique, dark green - ADAR1p150 unique, light green - ADAR1p110 unique with DDA and both ADAR1 isoforms with DIA), outline colors indicate the experiment in which the HCIs were identified and the thicknesses of the interconnected lines denote the combined scores for particular interactions. The clustering annotation was based on an enrichment analysis with the STRING app in Cytoscape. Distinct proximal protein networks of ADAR1 isoforms during homeostasis. (A) Scheme depicting the experimental approach used for the identification of proteins proximal to ADAR1. Doxycycline-inducible stable cell lines 293 T-REx expressing either ADAR1p110 (yellow) or ADAR1p150 (green) with BirA* at the N-terminus were generated, enabling isoform-specific proximity labeling. The biotinylated proteins bound to the beads were subjected to LC–MS/MS in the DDA mode. The image was created with BioRender. (B) Venn diagram depicting the extent of overlap between the high-confidence neighboring proteins for both ADAR1 isoforms identified with BioID. (C) Gene ontology (GO) terms of the cellular components significantly enriched for the proximal protein network of ADAR1p110 and ADAR1p150. The height of the bar corresponds to the fold enrichment and its color to the ADAR1 isoform. Small subunit (SSU) processome, cytoplasmic ribonucleoprotein granules (c. RNP granules), P-bodies (PBs), cytoplasmic SGs (c. SGs), coding region instability determinant (CRD)-mediated mRNA stability complex (CRD-complex). (D) Functional protein network of high-confidence proximal proteins identified with BioID for both ADAR1 isoforms. Node colors correspond to the ADAR1 isoforms to which the protein is proximal (red– not present, light red – decreased, grey – unchanged, blue - increased), line thickness indicates combined String score and the node size corresponds to bait-normalized spectral counts of the identified proximal protein. The clustering annotation is based on the enrichment analysis with the STRING app in Cytoscape and the additional comparison to a published dataset was performed for SGs and PBs. Effect of RNase A digestion on ADAR1 protein interactions. (A) Scheme depicting the experimental approach used for the identification of endogenous ADAR1 interactors occurring on long RNA. A549 cell lysates were incubated with RNase A or RNase inhibitor for 1 h at 4°C before adding either ADAR1 or IgG antibodies for subsequent overnight incubation. The image was created with BioRender. (B) Number of ADAR1 HCIs identified after SAINT filtering of co-IP samples with and without RNase A treatment. (C) The physical protein interaction network of HCIs identified in the RNase experiment and the two endogenous ADAR1 co-IP experiments. Node colors represent qualitative comparisons of bait-normalized spectral counts of HCIs between samples with and without RNase A treatment. Line thickness indicates combined Saint score, and the node size corresponds to bait-normalized spectral counts of the identified interactor. Effect of ADAR1 EAA mutants on protein interactions and networks. (A) Scheme depicting experimental approaches used for the identification of stable interactors and proximal protein networks occurring on dsRNA. Doxycycline-inducible stable cell lines (293 T-REx) expressing either ADAR1p110 (yellow) or ADAR1p150 (green) as well as ADAR1p110 EAA and ADAR1p150 EAA isoforms were generated. Both Strep-tag IP and BioID were performed and the bound proteins were subjected to LC–MS/MS analyses. The image was created with BioRender. (B) Number of HCIs or high-confidence proximal proteins identified with the WT and ADAR1 EAA mutants of both ADAR1 isoforms with the different methods. The color of the bar corresponds to the ADAR1 isoform (yellow– ADAR1p110 and green – ADAR1p150). (C) Dot plot of ADAR1 interactors or proximal proteins identified as showing changes with RNase A treatment that overlap in the two different experiments (Strep-tag IP and BioID) for both ADAR1 isoforms and the effect of the ADAR1 EAA mutant. The size of the dot corresponds to bait-normalized average spectra of the identified protein and color corresponds to the ADAR1 isoform (yellow– ADAR1p110 and green – ADAR1p150). The blue line beside the gene name of some proteins, marks those with nucleolar localization. (D) Dot plot of ADAR1 interactors or proximal proteins that are identified as showing no changes after RNase A treatment that overlap in the Strep-tag IP and BioID experiments for both ADAR1 isoforms and the effect of the ADAR1 EAA mutant compared to WT protein. The size of the dot corresponds to bait-normalized average spectra of the identified protein and color to ADAR1 isoform (yellow– ADAR1p110 and green – ADAR1p150). Blue line marks the proteins with nucleolar localization. Results The ADAR1 interactome in homeostasis comprises proteins with roles in RNA metabolism To identify stable high-affinity interactions of ADAR1, endogenous ADAR1 was co-immunoprecipitated (endo-IP) from A549 lung carcinoma epithelial cells (Figure 1A and Supplementary Figure S1A and B). However, with this approach, it is not possible to differentiate between interactions formed with the nuclear ADAR1p110 and those with the cytoplasmic ADAR1p150 isoform. To address this limitation, inducible Flp-In™ T-REx™ 293 cell lines (henceforth referred to as 293 T-REx) were generated to separately express different ADAR1 isoforms; ADAR1p150 or ADAR1p110, with Strep II-HA tags at the N-termini for Strep-tag IPs (Figure 1A and Supplementary Figure S1C) (41,42). The proteins enriched following Strep-tag IPs were subjected to two distinct MS/MS data acquisitions. The DDA method detects the most abundant sample components while the DIA method provides a more comprehensive coverage of interacting proteins, particularly the low-abundance components in the sample (43,44). The HCIs were determined with SAINT and CRAPome filtering for the endo-IP and LIMMA filtering cut-off was used for Strep-tag experiments, as described in the Materials and Methods section. The number of HCIs was comparable between the two endo-IP experiments and the Strep-tag IP with DIA, while the number of HCIs were lower for the DDA dataset (Supplementary Figure S1D and Supplementary Table S1). As the two endogenous ADAR1 co-IP experiments exhibited some distinct HCIs, we focused our subsequent analyses only on those that were common to both (Supplementary Table S1). A comparison of the endogenous ADAR1 co-IP and Strep-tag IP HCI datasets for both ADAR1 isoforms revealed that 9 and 37 common interactors for DDA and DIA respectively. (Figure 1B). An additional 11 HCIs of endogenous ADAR1 were identified as ADAR1p110-unique with DDA, while using DIA slightly increased the number to 17. A significant overlap in HCIs was observed between the two ADAR1 isoforms with DDA, with unique proteins interacting only with ADAR1p110 (Figure 1B). A similar overlapping pattern was observed with DIA, although this acquisition technique identified a relatively small number (21 in total) of ADAR1p150-unique interactors. Furthermore, an overlap in the interactomes of both isoforms is also evident at the MS1 or DIA intensity levels. (Supplementary Figure S1E). These results indicate that, in homeostasis, ADAR1p150 forms relatively few stable interactions, the majority of which are also observed with ADAR1p110, implying that they might represent interactions of the small fraction of ADAR1p150 protein localized in the nucleus (5,45). The HCIs identified as common to endogenous ADAR1 and Strep-tagged ADAR1 isoforms are likely to represent true stable, high-affinity ADAR1 interactors and potential regulators of A-to-I RNA editing, irrespective of cell type. The functional protein network of these common HCIs revealed three main clusters (Figure 1C). These clusters include proteins involved in rRNA biogenesis and processing, mRNA processing and a mixed cluster of endoribonuclease complexes and deadenylation-dependent mRNA decay proteins. The majority of proteins were ADAR1p110 interactors or were found with both ADAR1 isoforms. While SRRM2, MAP7, ZC3HAV1 and LARP1 interact solely with ADAR1p150. The majority of HCIs displayed in the network are localized in the nuclear and nucleolar compartments and their predominant identification is as HCIs of ADAR1p110. (Supplementary Figure S1F). In summary, ADAR1, in homeostasis, interacts with proteins involved in RNA metabolism, mainly related to rRNA biogenesis and processing and mRNA processing. Moreover, unique interactions are specific to the ADAR1p110 isoform, while ADAR1p150 shares some common interactors but forms relatively few stable interactions. Delineating proximal protein networks of ADAR1 isoforms with BioID Both endo-IP and Strep-tag IP methods revealed stable and high affinity interactors which were predominantly localized in the nucleolus and the nucleus. The majority of the interactors were found with both ADAR1 isoforms or were unique to the ADAR1p110 isoform, suggesting that ADAR1p150 interactions in the cytoplasm are more transient. To further address this question and identify proteins that transiently interact with ADAR1 we have utilized BioID, a proximity labeling method (46). Two ADAR1 isoforms were tagged with BirA* at the N-terminus and individual cell lines were created (Figure 2A). Following enrichment of biotinylated proteins (Supplementary Figure S2A) LC–MS/MS analysis in the DDA mode was performed. In contrast to Strep-tag IPs, the number of high-confidence proximal proteins (term used to describe the proteins identified with BioID), was slightly higher but comparable for both isoforms (Supplementary Figure S2B and Supplementary Table S2). Both ADAR1 isoforms displayed distinct proximal protein networks with 6 proteins being proximal to both isoforms (Figure 2B). These included, besides ADAR1, ILF3 (47,48), ZFR (48), DHX9 (49–51), known interactors of ADAR1, as well as STAU2 and RBM33. The GO-term enrichment analyses emphasize the differences in the proximal protein networks between ADARp110 and ADAR1p150 (Figures 2C, Supplementary Figure S2C, and Supplementary Table S2). ADAR1p110 proximal proteins are primarily found in the nucleoplasm, on the chromosomes, and in the nucleolus (SSU processome and 90S preribosome). In contrast, proteins interacting with ADAR1p150 are enriched for cytoplasm, cytoplasmic RNP granules, PBs, cytoplasmic SGs and CRD-mediated mRNA stability complex (CRD-complex). The individual high-confidence proximal proteins identified for each isoform were used to generate a functional protein network (Figure 2D). A clear distinction was observed between the ADAR1 isoform protein clusters, with few proteins being common. However, these proteins show different relative abundances, as indicated by the node sizes. Of these, ILF3, ZFR and DHX9 are more enriched with ADAR1p110 while STAU2 and RBM33 are similar between the two isoforms. Other high-confidence proximal proteins clustered in three different main clusters. Those proximal to ADAR1p110 included proteins with roles in rRNA processing and RNA helicase activity, a cluster also observed for stable interactions (Figure 1C). Also proximal to ADAR1p110 were components associated with the U2-type spliceosomal complex and mRNA splicing. The components of these two main clusters are in agreement with the localization of ADAR1p110 (3,4,45). The ADAR1p150 proximal protein network on the other hand was more heterogeneous. Considering that GO-terms enriched with ADAR1p150 are related to cellular condensates (Figure 2C), we compared the BioID dataset with the dataset of SGs and PBs components (52). The majority of ADAR1p150 proximal proteins are described as a component of SGs, PBs, or of both (Figure 2D, Supplementary Figure S2D and Supplementary Table S2). Considering that our cells were unstressed, finding significant overlap with these condensates indicates that ADAR1p150 could be a part of the core proteins required for the rapid formation of the condensates during stress (52,53). Moreover, recent profiling of RBPs binding dynamics during the mRNA life cycle uncovered that the SGs proteome is enriched on aged mRNAs (54). Therefore, we compared the proximal protein networks of both ADAR1 isoforms with this dataset. ADAR1p110 proximal RBPs display nuclear mRNA binding dynamic, while ADAR1p150 proximal RBPs have a broader range suggesting cytosolic mRNA binding (Supplementary Figure S2E and Supplementary Table S2). Moreover, the ADAR1p110 proximal proteins are found to belong to clusters II, III, and IV and ADAR1p150 proximal proteins to VI and VII (Supplementary Figure S2F). Clusters II, III and IV were enriched for pre-mRNA splicing, nucleocytoplasmic transport and rRNA processing while clusters VI and VII were enriched for translation regulation, SG, and mRNA stability. These terms were taken from the original publication (54). Lastly, we find a certain overlap of the BioID dataset with those we identified as stable interactor for both isoforms, with more proteins in common with the ADAR1p110 isoform (Supplementary Figure S2G and Supplementary Table S2). Three proteins are identified regardless of the method used including ILF3 which was present with both ADAR1 isoforms, while KIF23 and IGFBP1 were found with ADAR1p110 and ADAR1p150, respectively (Supplementary Figure S2H). Moreover, proximal proteins identified with ADAR1p110 show more overlap with the Strep-tag dataset while those identified with ADAR1p150 have a greater overlap with the endo-IP dataset. In summary, proximity labeling allows us to distinguish distinct ADAR1 isoform-specific proximal protein networks not observed with other methods. The protein networks and associated biological roles match the predominant localization of each isoform. The proximity of ADAR1p150 to core components of cellular condensates and to ‘aged’ mRNPs was also revealed. Impact of RNase treatment on endogenous ADAR1 interactions As ADAR1 is a dsRNA-binding protein, performing co-IP can enrich for interactors that occur on longer RNA molecules. To determine which of the identified HCIs of endogenous ADAR1 are bridged by RNA molecules, the A549 cell lysate was incubated with RNase A or left untreated prior to co-IP with endogenous ADAR1 and LC–MS/MS analysis (Figure 3A). RNase A treatment did not affect the immunoprecipitation of ADAR1 (Supplementary Figure S3A) but it resulted in significant RNA degradation (Supplementary Figure S3B). RNase A treatment resulted in fewer HCIs (Figure 3B and Supplementary Table S3) even though the total number of identified proteins was higher (Supplementary Figure S3C). Importantly, for this endo-IP experiment, we defined HCIs utilizing SAINT filtering while omitting CRApome filtering, in contrast to the other two endo-IP experiments described in Figure 1. This was due to differences in ADAR1 spectral counts in the three experiments which can affect CRAPome filtering. Nevertheless, we identified high number of proteins overlapping between the RNase A and the two endo-IP experiments (Supplementary Table S3). These overlapping HCIs were used to generate a protein network of physical interactions which revealed four main clusters (Figure 3C). Further comparison of ADAR1-normalized spectral counts of HCIs between samples with and without RNase A treatment revealed the effect of RNase A treatment on HCIs (Figure 3C). The more distal interactors of endogenous ADAR1 include proteins from all clusters; for example, various heterogeneous nuclear RNPs (hnRNPs) as well as some proteins with roles in ribosomal processing. However, not all hnRNPs seem to be distal interactors. A significant number of HCIs were unaffected and a few were even increased with RNase A digestion, indicating that these are likely to comprise of a group of close interactors. We have validated some of the interactions and the effect of RNase A treatment with endo-IP and immunoblotting (Supplementary Figure S3D). Quantification of the signal intensity agrees with results obtained with the LC–MS/MS data for all tested proteins except for NOP2, which was not significantly depleted with RNase A as determined with immunoblotting (Supplementary Figure S3E). Of HCIs unaffected by the RNase A treatment we identified some of the dsRBDs, including ILF2, ILF3, DHX9, STAU1, STAU2, TARBP2, PKR (EIF2AK2) and PRKRA, of which the majority are known ADAR1 interactors. While PKR, ILF3, and ILF2 were demonstrated to be ADAR1 interactors bridged by RNA molecules (16,47); conflicting results are found in the literature for DHX9 (17,49). Quantification of the signal obtained with endogenous ADAR1 co-IP and immunoblotting revealed a high degree of variability for some interactors with RNase A treatment (Supplementary Figure S3D and E). This could be also attributed to the incomplete digestion of RNA by RNase A and the presence of remaining RNA fragments (Supplementary Figure S3B and F). The presence of the RNA fragments which proteins may bind to, could lead to the false identification of the interactor as RNA-independent. By using RNase A digestion prior to endogenous ADAR1 co-IP and subsequent LC–MS/MS analysis, we identified proximal and distal interactomes of ADAR1 in A549 cells. Subsequent co-IP and immunoblotting experiments are in agreement with LC–MS/MS data, but they also revealed high variability. Importantly, some proximal interactions might occur on the remaining RNA fragments following the RNase A digestion. Effect of ADAR1 dsRNA-binding mutants on protein interactions and proximal protein networks An ADAR1 dsRNA-binding mutant was used to elucidate whether the protein interactions on residual RNA fragments that are unaffected or even increased by RNase A treatment, is due to incomplete RNA digestion or if they represent direct protein interactions. 293 T-REx cells were generated to express individual ADAR1 isoforms with mutations in the KKxxK motifs in all three dsRBDs (ADAR1 EAA) which are known to disrupt ADAR1 dsRNA binding (Figure 4A) (21). Individual ADAR1 EAA isoforms were tagged with Strep II-HA or BirA* at the N-terminus, allowing for the identification of HCIs following Strep-tag IP and also the identification of the proximal protein networks with BioID. Introducing ADAR1 EAA mutations resulted in complete losses of stable interactions for both isoforms, regardless of the data acquisition technique used (DDA or DIA) (Figure 4B). However, using BioID, ADAR1 EAA mutants had a slightly higher number of high-confidence proximal proteins, especially the ADAR1p110 isoform (Figure 4B). ADAR1 levels were determined by immunoblots for the respective epitope tags; after Strep-tag IP (Supplementary Figure S1C) or BioID (Supplementary Figure S2A) the level of ADAR1 EAA mutant isoforms were slightly lower than WT ADARs. This trend was also observed for ADAR1 spectral counts and with the Strep-tag DDA method, but not for BioID (Supplementary Figure S4A). Nevertheless, the numbers of identified bound proteins were comparable between WT proteins and EAA mutants for the Strep-tag method, with more proteins in the ADAR1p110 EAA BioID dataset (Supplementary Figure S4B and Supplementary Table S3). Further comparisons between WT and EAA revealed that only a third of ADAR1p110 EAA and half for ADAR1p150 EAA proximal proteins were also identified with WT protein (Supplementary Figure S4C and Supplementary Table S3). In addition, a high number of proteins were proximal to EAA only. This suggests significant changes in the proximal protein network due to the point mutations in dsRBDs. The consequences of mutations are evident in subtle changes in subcellular localization as observed by immunofluorescence for both the Strep-tagged (Supplementary Figure S4D) and BirA*-tagged (Supplementary Figure S4E) ADAR1 EAA isoforms, which supports the BioID results. The effect of the mutation is more prominent for the ADAR1p110 EAA isoform, as indicated by the more widespread localization in the nucleus and thus higher number of novel proximal proteins. The effect of ADAR1p110 EAA on individual interactors is depicted in a protein network (Supplementary Figure S4F), emphasizing the dominant effect this mutation has on the ADAR1p110 isoform. In contrast the protein network of ADAR1p150 EAA reveals that few proteins are affected (Supplementary Figure S4G). Only a few components of SGs and PBs in unstressed conditions are affected by the ADAR1 EAA mutations, while a comparison to temporal mRBP clusters classified in this study (54) revealed that half of the components from each cluster were lost with ADARp150 EAA (Supplementary Figure S4H). The ADAR1p110 EAA mutant displayed a different pattern as it affected the majority of components from overlapping clusters. We compared HCIs of endogenous ADAR1 in A549 cell after RNase A treatment (Figure 3) with the those identified with Strep-tag and BioID. This comparison was performed for both ADAR1 isoforms to compare the effects of RNase A digestion and ADAR1 EAA mutation (Figure 4C and D). The majority of HCIs that are not present or decreased following RNase A treatment are nucleolar proteins and are identified with ADAR1 WT and not with ADAR1 EAA, except for PDCD11 and DDX24, which validates the different methods used to identify this set of interactions (Figure 4C). The second group of interactions affected by RNase A digestion is predominantly found with the ADAR1p150 isoform, where the majority of proximal proteins are not affected by ADAR1p150 EAA, the exception being FXR1 and MTDH. This suggests that the dsRBD mutations may have a smaller effect on ADAR1p150 protein localization and thus are detected with ADAR1p150 EAA. This may indicate contributions of the catalytic deaminase domain or the Zα RNA-binding domain to ADAR1p150 EAA dsRNA-binding. With the exception of DHX9, all other HCIs unaffected by RNase A digestion were also identified as interactor or proximal proteins to ADAR1p110 and were not detected with the ADAR1p110 EAA mutant (Figure 4D). In contrast, those identified with ADAR1p150 were found proximal to both ADAR1p150 WT and ADAR1p150 EAA mutant, with ILF3 protein being the exception. In summary, utilizing the ADAR1 dsRNA-binding mutant (ADAR1 EAA) in combination with RNase A digestion offers a unique insight into direct protein–protein interactions and interactions bridged via dsRNA. However, our results indicate that introducing these mutations affects protein localization, and thus the proximal protein networks, especially for the ADAR1p110 isoform which provides information about the location of the protein interactions. Comparison of ADAR1 interactome with known interactors and RNA editing regulators The methods we used resulted in the identification of novel interactors for both isoforms of ADAR1 during homeostasis. When comparing our dataset with the BioGRID (55) and IntACT (56) databases, we identified some of the previously described ADAR1 interactors (Figure 5A and Supplementary Table S3). As expected, more overlapping proteins were observed with endo-IP dataset and with ADAR1p110 particularly with DIA acquisition. Two other studies performed large scale identification of ADAR1 protein network with proximity labelling. These included BioID of nuclear ADAR1 in HeLa and M17 cell (16) and APEX2 of ADAR1p110 in MCF-7 cells (17). Comparative analysis of the proximal protein network of ADAR1p110 WT and ADAR1p150 WT reveal proximal proteins which overlap with those identified predominantly with ADAR1p110 WT, is reproducible and irrespective of cell-type (Supplementary Figure S5A and Supplementary Table S3). Comparison of interactors identified with our methods to known interactors and regulators of RNA editing. (A) Number of HCIs or proximal proteins from different datasets and the number of proteins previously described as ADAR1 interactors in two databases (BioGRID and IntACT). The height of the bar corresponds to the number of HCIs, and the filled area indicated whether they are either known or novel. (B) Heatmap showing the identification of some known proteins that interact with ADAR1 and the requirement of RNA for the interaction. Information about the RNA-dependence of the interaction was extracted from the original publication (Supplementary Table S3) and is shown on the side (N.A., not available, RNA-indep., RNA-independent, Var., variable effect, RNA-dep., RNA-dependent). The color of the tile on the heatmap corresponds to the ADAR1 isoform and the effect of RNase A digestion or ADAR1 EAA mutant (yellow– ADAR1p110 WT and EAA, light yellow – ADAR1 WT and depleted with ADAR1p110 EAA, green – ADAR1p150 and EAA, light green - depleted with ADAR1p150 EAA). Bar graphs on the right panel of the plot indicate the effect of depletion of the identified interactor on ADAR editing activity. The length of the bar represents the percentage of differentially edited sites, and the color corresponds to the direction of changes in editing (red– increased, light yellow – unchanged, blue – reduced). Analysis of published interactors and the effect of RNase treatment on the interactions was also performed (Figure 5B and Supplementary Table S3). As previously observed, RNase A digestion prior to endo-IP gave reproducible results for NCL and HNRNPL, which are described as RNA-dependent interactors. However, this was not observed for ILF3, ILF2, STRBP and PRKRA, while STAU1 and PKR, which show variable effects in the literature, are present with endogenous ADAR1 even with RNase A digestion but were lost or not identified with the EAA mutant. This suggests the need for two different approaches to elucidate true protein–protein interactions. Moreover, several studies have investigated regulators of A-to-I editing by ADARs; however, Quinones-Valdez et al. (57) used a more global approach to understand the effect of RNA-binding proteins on ADAR editing in both K562 and HepG2 cell lines. We compared the interactors they identified with our dataset to further correlate whether the interaction may affect RNA editing (Figure 5B, Supplementary Figure S5B and Supplementary Table S3). Interactors found to be common with the different methods, and revealed that depletion of these RBPs had a variable effect on editing in both cell lines tested. Similar to the published observation, the majority of HCIs or high-confidence proximal proteins which we identified as overlapping with the published study, have induced differential editing in <10% of the associated testable editing sites after their knockdown in K562 or HepG2 cells (Supplementary Figure S5B). However, a higher number of identified RBPs increased editing when depleted, suggesting that these identified interactors are novel negative regulator of ADAR1 editing activity. Together, these findings suggest that our dataset largely aligns with the published data, but it also provides new insights into ADAR1 interactions and novel cell specific editing regulators. Identification of novel ADAR1 interactors during IFN induction without significant changes to the core ADAR1 interactome Cytoplasmic ADAR1p150 is regulated by IFN and this is crucial to prevent aberrant immune responses to self-dsRNA (7,18); however, currently there is no information on the ADAR1 interactome during the IFN response. To address this, we extended our analysis of the ADAR1 interactome by analyzing the late IFN response. For this purpose, A549 cells were treated with IFNα or p(I:C), which is a synthetic dsRNA that mimics a viral infection, for 30 h before co-IP with endogenous ADAR1, while 293 T-REx cells expressing individual ADAR1 isoforms were treated with IFNα for 24 h before Strep-tag IP (Figure 6A). These treatment conditions resulted in induction of the IFN response, confirmed by increased expression of RIG-I, MDA5 and PKR proteins (Supplementary Figure S6A and B). Moreover, treatment of 293 T-REx cells with IFNα also upregulated the expression of individual endogenous retroviruses localized in the 3′UTRs of TRIM22 (MLT1C49), OAS2 (MLT1C627) and DDX60L (MLTA10) (Supplementary Figure S6C). Levels of endogenous ADAR1 following co-IP from A549 cells were similar between NT and as expected IFNα-treated cells expressed more ADAR1p150 after IFN treatment (Supplementary Figure S1B). Levels of Strep-tagged ADAR1 were comparable across all samples (Supplementary Figure S1C). Interactors of ADAR1 during the IFN response. (A) Scheme depicting treatments used to elucidate ADAR1 interacting protein networks during the IFN response. A549 cells were treated with either IFNα or p(I:C) for 30 h before endogenous ADAR1 co-IP. 293 T-REx cells expressing ADAR1 isoforms tagged with Strep-tag were treated with IFNα for 24 h before collection. This image was created with BioRender. (B) Number of HCIs identified as IFN-related across different treatments and datasets from this study. The height of the bar corresponds to the number of HCIs, and the gray-filled area shows those that are identified as IFN-related, with the exact number shown above the highlighted area. Proteins were annotated with the Interferome database and with GO-terms for biological processes related to IFN response. (C) Dot plot of ADAR1 interactors having log2FC (treatment over NT samples) >1 for Endo-IP, Strep-tag IP (DDA and DIA). Only HCIs identified in all three endogenous ADAR1 co-IP experiments are shown. Both the size of the point and the color gradient correspond to the log2FC of protein enriched during IFN response. (D) Dot plot showing the effect of RNase A treatment on ADAR1 interactors shown in panel (C) that are enriched with ADAR1 during the IFN response. The size of the point corresponds to the log2FC of protein enriched during the IFN response and the color to the direction of the change of the interactor upon RNase A treatment (pink- not found, grey – unchanged, blue - increased). Initially, we defined the proteins in our datasets as IFN-related using the following criteria: >2-fold changes in the Interferome database (36) and the presence of terms; IFN, immune response and viral regulation under biological processes in Uniprot. With these criteria we found that, in each method we used, a certain number of HCIs were IFN-related (Figure 6B and Supplementary Table S4). However, not all identified IFN-related proteins were HCIs (Supplementary Figure S6D). The elevated number of IFN-related proteins in NT conditions can be attributed to the basal level of IFN response previously observed in the A549 lung cancer cells (15) and evidenced by the presence of the ADAR1p150 isoform in NT conditions following endogenous ADAR1 co-IP (Supplementary Figure S1B). To overcome this and to identify increased interactions during the IFN response, imputed and bait-normalized MS1 intensity values of identified interactors were used for comparison between IFN-treated and NT conditions. HCIs enriched over NT conditions (log2FC higher than one) were predominantly IFN-related (Figure 6C and Supplementary Table S4). Three independent endogenous ADAR1 co-IP experiments two with IFNα and one with p(I:C) treatment for 30 h, identified mainly regulators of the IFN response. However, of these seven HCIs, three were found to also overlap with those identified with Strep-tag co-IP in 293 T-REx. The DDA analysis identified only HERC5 as an HCI of both ADAR1 isoforms. HERC5 is a polysome-associated ISG15 ligase that modifies newly synthesized host and viral proteins during the IFN response (58,59). In agreement with our data, ADAR1 was previously described to undergo ISGylation during the IFN response, but the ISG-15 ligase has not been previously described (60). Two additional HCIs: HELZ2 and OAS3, were identified in 293 T-REx only with the DIA. The difference in identification of HCIs between endogenous ADAR1 co-IP from A549 cells and Strep-tag IP from HEK293T cells is probably due to the different levels of expression of IFN-related proteins observed in the two cell types (Supplementary Figure S6E). Interestingly, HERC5 and HELZ2 were identified as HCIs of both ADAR1 isoforms whereas OAS3 interacts only with ADAR1p150. HELZ2, which is an IFN-inducible 3′-5′ exoribonuclease and a helicase, is active on structured RNAs and has roles in regulating LINE-1 elements and in the IFN response (61,62). Moreover, both HELZ2 and HERC5 recognize LINE-1 RNA, affecting retrotransposition (62). Similarly, ADARs were also described to affect LINE-1 retrotransposition, however, this effect was tested in homeostasis with ectopically expressed LINE-1 and ADARs (63). OAS3 is a member of the 2′-5′ oligoadenylate synthetase (OAS) family activating RNase L; it requires long dsRNA for its activation, displays high-affinity towards dsRNA (64–66). Crosstalk between ADAR1 and the RNaseL pathway has been described in A549 cells, but the underlying mechanism of inhibition remains elusive (67). We utilize AlphaFold Multimer (AFM) and predicted a direct interaction between the ADAR1 deaminase domain of ADAR1 and OAS domain I of OAS3 (Supplementary Figure S6F). The interacting residues of ADAR1 are within the conserved dimerization helix identified as responsible for ADAR2 dimerization (68). This suggests a novel protein–protein interaction between ADAR1 and OAS3 as an underlying mechanism for inhibition but this requires further validation. To elucidate if the interactions we identified during the IFN response can occur on longer RNA, RNase A digestion was performed prior to endo-IP in A549 cells, as described in Figure 3 during homeostasis. Using SAINT analysis, HCIs were further filtered for those enriched during IFN induction and for those common for all three endo-IP experiments (Figure 6D). DTX3L is identified as a HCIs only in samples without RNase A treatment, indicating that their interaction with ADAR1 is bridged by longer RNA (Figure 6D). While bait-normalized spectral counts of OAS3 protein are similar before and after RNase A digestion, other HCIs are slightly increased with RNase A digestion. Such interactors could occur independently of RNA substrate or be present on the shorter RNA fragments generated during RNase A digestion, as described previously in Figure 3. Therefore, we have also performed Strep-tag IP of ADAR1 EAA mutants, dsRNA-binding mutants previously described in Figure 4, following IFNα treatment. However, similarly to results obtained for the ADAR1 EAA mutants in homeostasis, no HCIs were identified for Strep-tag IP regardless of LC–MS/MS acquisition (DDA or DIA) (Supplementary Figure S6G). These results indicate that interactions of ADAR1 isoforms with HERC5, HELZ2 and OAS3 require ADAR1 to bind to dsRNA. To summarize, ADAR1 forms previously undescribed stable interactions with IFN-related proteins during IFN induction in both A549 and 293 T-REx cells to varying levels. These interactions require ADAR1 binding to dsRNA. A short HMW poly(I:C) treatment during the IFN response induces changes in the proximal protein network of both ADAR1 isoforms We have shown that during an extended treatment with IFNα or p(I:C), ADAR1 forms novel interactions with regulators of the antiviral response. However, the experimental approach is restricted to stable interactions that occur during the later stages of the IFN response and in the absence of exogenous dsRNA. To address this, samples were treated with IFNα for 24 h, together with samples where cells in addition to IFNα were transfected with p(I:C) for 4 h before collection (Figure 7A). We observed the proximity of widespread cytoplasmic staining of ADAR1p150 and G3BP1 in NT and IFNα-treated conditions (Figure 7B). This is in agreement with our observation that ADAR1p150 is proximal to the core of the SGs in unstressed conditions, as shown in Figure 2. The presence of p(I:C) during the IFN response resulted in G3BP1 positive foci being formed. G3BP1 is a marker for SG formation (69,70). ADAR1p150 has been previously shown to co-localize with the G3BP1 during oxidative stress and treatment with p(I:C) (71). We observed different patterns of co-localization of Strep II-HA-tagged ADAR1p150 WT and endogenous G3BP1 in the presence of p(I:C) during IFN response. In some foci we observed ADAR1 signal overlapping with G3BP1, and we also observe ADAR1 in close proximity to G3BP1 while we also observe ADAR1 foci that has no overlapping signal with G3BP1. This was of interest considering that various forms of cytoplasmic condensates or SG-like condensates have been described in the literature (72) with ADAR1 being identified as a component in some of these. Changes in ADAR1 proximal protein networks in the presence of both IFNα and p(I:C). (A) Scheme depicting treatments used for elucidating the proximal protein network. 293 T-REx cells were treated with IFNα or IFNα + p(I:C). Total time of treatment was 24 h. This image was created with BioRender. (B) Proximal localization of the Strep II-HA tagged ADAR1p150 isoform (green) with SGs marker – G3BP1 (red) in merged images under three conditions; NT, IFNα and IFNα + p(I:C). DAPI is shown in blue. Scale bar is 10 μm (C) Venn diagram of the overlap of high-confidence proximal proteins identified in the three conditions tested for ADAR1p110 and ADAR1p150. The color corresponds to individual treatments. (D) Dot plot of ADAR1 proximal proteins identified as unique for IFN and IFN + p(I:C) with roles in the immune response. ADAR1p110 WT and ADAR1p110 EAA mutants with different treatments are shown on the left and ADAR1p150 proteins are on the right. The size of the dot corresponds to bait-normalized average spectra of the identified protein and the color corresponds to the treatment in which the protein was identified (pink – IFNα and purple – IFNα + p(I:C)). To elucidate ADAR1 protein dynamic during IFN response and in the presence of exogenous dsRNA, we treated 293 T-REx expressing BirA*-tagged ADAR1 isoforms and GFP control and performed BioID. This treatment resulted in the induction of IFN response as observed by the increased expression of RIG-I and PKR (Supplementary Figure S7A). The amount of biotinylation proteins were comparable between conditions (Supplementary Figure S2A) while high-confidence proximal proteins were slightly higher for ADAR1p110 in the presence of p(I:C) (Supplementary Figure S7B). A comparison of high-confidence proximal proteins illustrates that 57 proteins identified with ADAR1p110 and 74 with ADAR1p150 are common for all conditions, indicating that a core protein network remained unchanged with these two treatments (Figure 7C and Supplementary Table S4). While 19 proteins proximal to ADAR1p150 are observed with IFNα treatment, the addition of p(I:C) resulted in most of the proteins overlapping with the IFNα treatment. In contrast, with ADAR1p110, IFNα treatment resulted in only a few uniquely identified proteins, while the addition of p(I:C) drastically increased the numbers of unique proximal proteins. To understand the changes in the proximal protein network during IFNα and p(I:C) treatments, the individual proteins identified uniquely under these conditions are indicated by different colors on the venn diagram in Figure 7C. These include proteins with different cellular localizations and some have known roles during the IFN response (Supplementary Figure S7C). We further filtered for proteins with described role in immunity as well as known ADAR1 interactors (Figure 7D). We observed that treatment with IFNα resulted in ADAR1p110 being proximal to XPO5 and STAU1, indicating relocalization of this isoform from the nucleus to the cytoplasm. This was not observed with the ADAR1p110 EAA mutant, indicating it was dependent on ADAR1p110 dsRNA binding. The export of ADAR1p110 from the nucleus under different stress conditions has been previously described and suggested to be dependent on dsRNA-binding (73–75). However, it has not been known to occur during IFN induction. While XPO5 was also proximal protein with p(I:C) treatment, STAU1 was not detected under this condition. Additional proteins proximal to ADAR1p110 WT but not to the ADAR1p110 EAA mutant were unique to the p(I:C) treatment included PKR (EIF2AK2), HERC5, SSB and PRMT1 which have roles in the IFN response and TNPO1, a known ADAR1 nuclear import receptor (74,76). These may represent novel nearby proteins and interactions occurring on exogenous dsRNA which in this case is p(I:C). As previously mentioned, the proximal protein network of ADAR1p150 was similar between IFNα alone and with the addition of p(I:C). However, a certain distinction exists between ADAR1p150 WT and ADAR1p150 EAA mutant proteins (Figure 7D). Proteins with a role in immune responses dependent on ADAR1p150 binding to dsRNA include NFX1, SRPK2, TRIP6, IRAK1 and TRIM25. This was observed for USP10 and ZC3HAV1 during IFNα treatment and the presence of p(I:C) however as these two proteins were also proximal to ADAR1p150 EAA mutant their interaction is independent of binding to RNA. TTLL12 protein was proximal to both ADAR1p150 WT and ADAR1p150 EAA only with p(I:C). Identification of TRIM25 and ZC3HAV1 suggests that under these conditions ADAR1p150 is proximal to some of the components of antiviral SGs (77). TRIM25 protein has been described as an important component of these granules triggered by transfection of p(I:C) and required for the regulation of antiviral responses via ubiquitination (77). Considering that we did not observe more ‘drastic’ changes in ADAR1p150 proximal protein network with p(I:C), the condensation of ADAR1p150 and G3BP1 as shown in Figure 7C, indicates the proximity of ADAR1p150 to the core of the SGs so it is readily available for formation of larger and microscopically visible condensates upon stress. Alternatively, formation of microscopically visible G3BP1 and ADAR1p150 condensates only occurs in some cells which limits the identification of proximal proteins with BioID. Discussion Here we employed proteomics approach to elucidate the ADAR1 interactome and proximal protein network providing a comprehensive resource for ADAR1 research. First, we performed co-IP with the endogenous ADAR1 protein from A549 cells. Next, we generated 293 T-REx cells expressing individual ADAR1 isoforms, tagging them with Strep II-HA for Strep-tag IPs to identify stable, isoform-specific, high-affinity interactions. In addition, we used BirA*-tagged ADAR1 isoforms for the BioID method to map the proximal protein network. To differentiate between proximal and distal interactions potentially mediated by RNA, we performed RNase A treatment prior to endogenous ADAR1 co-IPs and utilized the dsRNA-binding mutant (ADAR1 EAA) in the Strep-tag IPs and BioID experiments. Previously, two studies have performed large-scale approaches to explore the proximal protein network of ADAR1p110 isoform (17) and ADAR1 nuclear protein network during homeostasis (16). However, a significant gap in understanding the interactome of the ADAR1p150 isoform and the dynamics of ADAR1 interactions during stress responses, including IFN induction, remains. Therefore, to mimic the cellular conditions when ADAR1p150 is predominantly expressed, we treated cells with IFNα and mimicked viral infection with a short exposure to HWM poly(I:C). With this, we observed the dynamics in protein network of both ADAR1 isoforms under stress conditions with the BioID. Our approach identified novel ADAR1 interactors, in addition to those previously described that were dependent on both the ADAR1 isoform and the stress condition used. Stable HCIs common to both endogenous ADAR1 in A549 cells and Strep-tagged ADAR1 isoforms expressed in 293 T-REx cells included proteins involved in rRNA biogenesis and mRNA processing. Our results are consistent with the known ADAR1 isoform subcellular localizations and biological functions (3,4,74). Analysis of proteins identified with Strep-tag IPs with DIA, revealed additional HCIs that overlapped with the endogenous ADAR1 co-IP. They belong to the same clusters as those identified with DDA, indicating these interactors are of lower abundance and not cell-type specific. In addition, we observed that some stable interactions identified with Strep-tag IPs were unique to ADAR1p110, which is consistent with the nuclear localization of these proteins. However, some of the HCIs were common to both isoforms, suggesting that the ADAR1p150 interactome resembles that of ADAR1p110 during homeostasis. ILF3 was identified as a common interactor and other proteins such as FXR1, G3BP2 and HNRNPM were also found to interact with both isoforms. In contrast to stable interactors, BioID proximity labeling gave distinct protein networks for both ADAR1 isoforms. In homeostasis, the ADAR1p150 cytoplasmic protein network appears to consist of weaker or more dynamic interactions, which may explain why interactions with ADAR1p150 were not identified with the stable Strep-tag IP. This proximal protein network formed a heterogeneous cluster as observed with GO-terms enrichment analysis; however, further classification revealed an overlap primarily with SG components and proteins localizing to both SGs and PBs under conditions of no stress (52,53). This suggests that ADAR1p150 associates with core SG components in the absence of external stress. Indeed, co-localization of ADAR1p150 with subset of SG markers has been previously observed during arsenate or poly(I:C) treatment in HeLa cells (71). Alternatively, ADAR1p150 may associate with RNPs formed in later stages of the life of an mRNA as a significant number of ADAR1p150 proximal proteins belong to these later clusters associated with translation regulation, SG and mRNA stability terms (54). However, SG and aged mRNA clusters are not mutually exclusive as many proteins are present in both. The proximal protein network of ADAR1p110 identified with BioID comprises of a group of distinct proteins. They are similar to those present with endogenous ADAR1 co-IP and Strep-tag IP experiments and belong to clusters involved in ribosomal RNA (rRNA) biogenesis and processing. Spliceosomal components and certain individual proteins were observed to be unique to the BioID method. While the role of ADAR1 p110 in splicing has been investigated (78–80), its function in the nucleolus and rRNA biogenesis and processing, remains largely unexplored. Due to the lack of detected editing of rRNA by ADAR1, it is hypothesized that the nucleolus may serve as a storage compartment for the enzyme allowing for quick export to the nucleoplasm with the presence of the editing substrate (3). Moreover, this further explains subtle differences in proteins identified in both the Strep-tag and BioID datasets. To distinguish between proximal and distal interactions mediated by RNA, we performed RNase A digestion prior to co-IP of endogenous ADAR1 from A549 cell extracts. We detected short RNA fragments remaining in the RNase digested samples which may explain the significant number of interactors that remained unchanged after RNase A treatment. The possibility exists that some interactions deemed RNA-independent may instead occur on these residual small RNA fragments, leading to the misinterpretation of them as direct protein–protein interactions. For instance, our data indicates that some dsRBPs such as ILF3, ILF2 and PKR are RNA-independent interactors following RNase A digestion. However, previous studies have found that their interactions with ADAR1 are bridged by RNA (16,47). These studies used overexpression systems, whereas our experiments utilized endogenous ADAR1 co-IPs, which may account for the discrepancies. However, interactors such as HNRNPL, NCL, and DDX54 are lost with RNase A treatment in agreement with published findings (17,63). To address the potential influence of residual RNA fragments after RNase A treatment, we used the ADAR1 EAA mutant that contains point mutations in all three dsRBDs, that abolish ADAR1-dsRNA interactions through the dsRBDs. We observed that certain interactors previously identified as RNA-independent were lost with this mutant, while others remained, consistent with the RNase A digestion results. Both ADAR1 EAA isoforms lost their high-affinity stable interactors, as determined by Strep-tag co-IPs, and this correlates with changes in their proximal protein networks observed with BioID. The effect was more pronounced for ADAR1p110 EAA due to the loss of some nucleolar interactors, which might be attributed to the observed relocalization of ADAR1p110 EAA from the nucleolus to nucleoplasm. ADAR1p150 EAA showed more subtle changes; ADAR1p150 EAA localizes to the cytoplasm exclusively, confirming the true ADAR1p150 proximal protein network and excluding associations that might occur solely due to some common nuclear localization with ADAR1p110. In summary, digestion with RNase A can give variable results likely due to incomplete digestion of dsRNA. On the other hand, ADAR1 EAA mutants that are unable to bind dsRNA give less ambiguous results; however, as their cellular localization is altered this impacts their proximal protein network. A significant overlap of our datasets with the published interacting proteins was observed mainly for the nuclear ADAR1p110 isoform due to the lack of information on the interactome of ADAR1p150. Combining RNase A digestion and the dsRNA-binding mutant of ADAR1 illustrated the dependence on RNA of ADAR1 interactions described in the literature and classify interactions either as distal or proximal. Also, the comparison of datasets of RBPs acting as editing regulators, we observed that those present across different datasets are predominantly enriched for RBPs acting as negative regulators as their depletion results in an increase in RNA editing. For example, we identified SRSF5 which may represent a novel regulator of ADAR1 editing as it is analogous to SRSF9 which is a negative regulator of ADAR2 activity (81). However, observing that some interactors had no effect on ADAR1 editing activity, this could stem from editing-independent mechanism of ADAR1. ADAR1p150 is an IFN-inducible isoform that is essential for modulating innate immune responses. Despite this, protein interactions of ADAR1p150 and of nuclear ADAR1p110 have not been fully characterized during the IFN response. To address this, we investigated the ADAR1 interactome following IFNα treatment. IFN induction led to the identification of novel ADAR1 interactors without altering core interactomes. These novel interactors are involved in regulating IFN responses and remained unaffected by RNase A treatment, suggesting direct proximal protein–protein interactions rather than being present together on long RNA. However, these interactions still require ADAR1 dsRNA-binding as they were lost with ADAR1 EAA. Among the identified interactors during the IFN response HERC5, HELZ2 and OAS3 which emerged as definite interactors as they occur in A549 and 293 T-REx cells. Identification of HERC5, an E3 ISG-15 ligase is in agreement with the report of ISGylation of ADAR1 during the IFN response (60). The second interactor, HELZ2, is an IFN-inducible 3′-5′ exoribonuclease that degrades structured RNA and affects the IFN response (61,62). Moreover, both HERC5 and HELZ2 have been reported to bind LINE-1 RNA and inhibit LINE-1 retrotransposition (62). This suggests that their interaction with ADAR1 may occur on LINE-1 RNA during the IFN response, potentially influencing LINE-1 activity. ADAR1 has been reported to regulate LINE-1 retrotransposition under homeostatic conditions with overexpressed components but the mechanism still remains unclear (63). The control of LINE-1 is critical, as higher RNA levels can enhance the IFN response further (82–86). Our findings suggest that ADAR1 may influence LINE-1 activity during the IFN response through interactions with HERC5 and HELZ2. However, these proteins also have functions outside of the LINE-1 life cycle as HELZ2 binds to structured RNA, so the interactions with ADAR1 may occur on these RNAs. A third definite interactor we identified was OAS3, a member of the OAS family that activates RNase L upon sensing dsRNA (65,66). ADAR1 has been reported to play a role in directly modulating the OAS/RNase L pathway in cultured cells (67). OAS3 has not been identified as a LINE-1 interactor or regulator (62), suggesting that the interaction with ADAR1 may occur on a different endogenous dsRNA. So, the question remains what are the endogenous dsRNAs on which ADAR1 interacts with these other proteins after IFN induction? The interactions described above were identified during the later stages of IFN treatment when the levels of IFN-induced proteins were highest and they represent stable interactions. To gain insight into the temporal and spatial dynamics of ADAR1 during the IFN response and the consequences of exogenous dsRNA, cells were exposed to a short treatment with p(I:C) after IFN treatment, mimicking a viral infection. This resulted in G3BP1 condensates forming in some cells, which is indicative of SG assembly. The presence of p(I:C) resulted in changes primarily in the proximal protein network of ADAR1p110 which were not observed with the ADAR1p110 EAA mutant, indicating that dsRNA binding is required for these associations. We observed some proximal proteins which indicate the relocation of ADAR1p110 to the cytoplasm. Previously, the transport of ADAR1p110 to the cytoplasm was reported to occurred during ultraviolet stress; ADAR1p110 modulates apoptosis by competing with STAU1 for substrate occupancy (75). Our data suggest that a closer interaction between ADAR1p110 and STAU1 occurs during the IFN response. We also found that ADAR1p110 treated with p(I:C) and IFN, is associated closely with PRMT1, HERC5, and PKR (EIF2AK2). PRMT1 may serve as a novel post-translational regulator of ADAR1, as methylation sites have been predicted at Arg1090 in the deaminase domain and at the N-terminus of ADAR1 (87). While it is primarily ADAR1p150 that inhibits PKR activation (13,14) and we identified an interaction between ADAR1p150 and PKR during poly(I:C) treatment, its classification as an HCI was limited due to detection in only two out of three biological replicates. The interaction between ADAR1p110 and PKR is supported with reports that both ADAR1 isoforms can interact with PKR (14). ADAR1p150 exhibited less pronounced changes than ADAR1p110 in its proximal protein network during p(I:C) and IFN treatments as it retains its cytoplasmic localization (88). Following poly(I:C) transfection, additional proteins such as TRIM25 and ZC3HAV1 were identified as proximal to ADAR1p150. TRIM25 is an RNA-binding protein that is essential for the innate immune response as it ubiquitinates various sensors (77,89,90). The loss of TRIM25 interaction with ADAR1p150 EAA suggests that ADAR1 proximity is dsRNA-dependent. ZC3HAV1 inhibits various viruses by binding viral RNA and promoting translational repression (88); its localization to SG is essential for its antiviral function (91). These proteins are known to localize to SG formed during viral infection or poly(I:C) treatment, which are named antiviral SGs (77). This implies that poly(I:C) may promote ADAR1p150 association with SG-like granules, which are distinct from those that form as a response to oxidative stress. Various granules that are distinct from SGs have also been reported, including RNase L-dependent bodies (92), dsRNA-induced condensates of antiviral proteins (93) and antiviral SGs (77,94). Our results indicate that ADAR1p150 association with SG-like granules may depend on the type of stress or RNA present. Our study demonstrates that the network of protein interactions involving different ADAR1 isoforms is complex, multilayered and dynamic under both homeostatic conditions and during the IFN response. These findings open multiple avenues for further research, which will not only clarify how ADAR1 regulates the IFN response but also how ADAR1 itself is regulated during this response. Supplementary Material Acknowledgements CIISB, Instruct-CZ Centre of Instruct-ERIC EU consortium, funded by MEYS CR infrastructure project LM2023042, is gratefully acknowledged for the financial support of the measurements at the CEITEC Proteomics Core Facility. Computational resources were provided by the e-INFRA CZ project (ID:90254), supported by MEYS CR. We acknowledge the core facility CELLIM supported by the Czech-BioImaging large RI project (LM2023050 funded by MEYS CR) for their support with obtaining scientific data presented in this paper. Molecular graphics and analyses performed with UCSF ChimeraX, developed by the Resource for Biocomputing, Visualization and Informatics at the University of California, San Francisco, with support from National Institutes of Health R01-GM129325 and the Office of Cyber Infrastructure and Computational Biology, National Institute of Allergy and Infectious Diseases. We thank Jiří Sedmík, Jaclyn Elizabeth Quin and Stanislav Stejskal for their valuable insightful discussions. Erik Képeš is kindly acknowledged for assistance with the R scripts and Valentina Lacovich Strašil for assistance with the immunofluorescence image acquisition and processing. Author contributions: Conceptualization: D.V. and M.A.O. Methodology: D.V. and M.A.O. Investigation: D.V., A.C., S.K., Y.T.B, M.M. and L.Y. Visualization: D.V. Funding acquisition and supervision: M.V. and M.A.O. Writing (original draft): D.V. Writing (review and editing): A.C., S.K., L.K., M.V. and M.A.O. Contributor Information Dragana Vukić, Central European Institute of Technology (CEITEC), Masaryk University, Kamenice 5, Brno 62500, Czechia; NationalCentre for Biomolecular Research, Faculty of Science, Masaryk University, Kamenice 5, Brno 62500, Czechia. Anna Cherian, Central European Institute of Technology (CEITEC), Masaryk University, Kamenice 5, Brno 62500, Czechia; NationalCentre for Biomolecular Research, Faculty of Science, Masaryk University, Kamenice 5, Brno 62500, Czechia. Salla Keskitalo, Institute of Biotechnology, HelsinkiInstitute of Life Science (HiLIFE), University of Helsinki, Helsinki 00014, Finland. Yih Tyng Bong, Institute of Biotechnology, HelsinkiInstitute of Life Science (HiLIFE), University of Helsinki, Helsinki 00014, Finland. Martin Marônek, Central European Institute of Technology (CEITEC), Masaryk University, Kamenice 5, Brno 62500, Czechia. Leena Yadav, Institute of Biotechnology, HelsinkiInstitute of Life Science (HiLIFE), University of Helsinki, Helsinki 00014, Finland. Liam P Keegan, Central European Institute of Technology (CEITEC), Masaryk University, Kamenice 5, Brno 62500, Czechia. Markku Varjosalo, Institute of Biotechnology, HelsinkiInstitute of Life Science (HiLIFE), University of Helsinki, Helsinki 00014, Finland. Mary A O’Connell, Central European Institute of Technology (CEITEC), Masaryk University, Kamenice 5, Brno 62500, Czechia. Data availability The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (95) partner repository with the dataset identifier: PXD049991, PXD049681, PXD049684, PXD049689 and PXD056080. Supplementary data Supplementary Data are available at NAR Online. Funding Grantová agentura České republiky (GAČR) [GAČR 20-11101S to D.V., GAČR by 22-26278K to M.M. and GAČR 21-27329X to M.A.O]. Conflict of interest statement. None declared. References Associated Data Supplementary Materials Data Availability Statement The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (95) partner repository with the dataset identifier: PXD049991, PXD049681, PXD049684, PXD049689 and PXD056080.