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

Diversification of the neuronal transcriptome through co- and post-transcriptional events is crucial for proper function of the mammalian brain. One of the most common modifications of neuronal transcripts is deamination of adenosine to inosine (A-to-I) in double-stranded RNA (dsRNA), which is catalyzed by one of two active adenosine deaminases that act on RNA (ADARs), ADAR1 and ADAR2 (also known as ADAR and ADARB1, respectively) (Bass, 2002; Bass et al., 1997; Nishikura, 2010; Paul and Bass, 1998). Most ADAR substrates depend on intronic sequences to form the dsRNA structure required for editing (reviewed in Wahlstedt and Öhman, 2011). Thus, editing of many substrates has to occur in the nucleus before splicing. Both ADAR1 and ADAR2 have been shown to accumulate in the nucleolus, where the enzymes bind but do not edit ribosomal RNA (Desterro et al., 2003; Sansam et al., 2003). However, the nucleolar association is highly dynamic and the ADAR enzymes shuttle between the nucleolus and nucleoplasm.

Owing to structural similarities, inosine is recognized as guanosine (G) by the cellular machineries. Hence, A-to-I editing has the capacity to diversify neuronal gene expression through alteration of protein sequences, splicing patterns and base-pairing properties. Editing within coding sequences mostly occurs in genes involved in neurotransmission, such as in the transcript coding for the GluA2 subunit of the AMPA glutamate receptor, where editing is essential for survival (Brusa et al., 1995; Higuchi et al., 2000). Another example of a recoding editing event occurs in the Gabra-3 transcript, encoding the α3 subunit of the inhibitory GABA receptor, where editing leads to an isoleucine to methionine (I/M) change in the translated protein (Ohlson et al., 2007). Here, the edited isoform is necessary for proper trafficking and recycling properties of the α3 subunit during brain development (Daniel et al., 2011). Most sites edited within coding sequences are inefficiently edited in the embryonic brain, while they are edited to much greater extents in the adult brain (Dillman et al., 2013; Hwang et al., 2016; Lomeli et al., 1994; Wahlstedt et al., 2009).

Just like edited sites within coding sequences, editing within repetitive elements in introns and untranslated regions (UTRs) is lower in embryonic tissue than in adult human tissue, including the brain (Shtrichman et al., 2012). Although most A-to-I editing sites within these repetitive elements have not been assigned specific functions, there are examples where editing leads to alternative splicing and exonization of introns (Lev-Maor et al., 2007). The non-coding microRNAs (miRNAs) are also edited, particularly in the targeting seed sequence (Alon et al., 2012; Ekdahl et al., 2012; Kawahara et al., 2007; Vesely et al., 2014). The ADAR enzymes target the nuclear miRNA precursors and can affect their maturation or mRNA-targeting capacity. The frequency of miRNA editing is also a regulated event during brain development (Ekdahl et al., 2012).

To date, mechanisms for how A-to-I editing is regulated have been sparse. We have previously seen that the protein levels of ADAR1 and ADAR2 are constant during mouse brain development, not following the gradual increase of A-to-I editing (Wahlstedt et al., 2009). A number of editing enhancer and repressor candidates have been identified by the Jantsch laboratory (Garncarz et al., 2013; Tariq et al., 2013), and work from the O'Connell laboratory has revealed that ADAR2 is post-translationally regulated by the phosphorylation-dependent peptidyl-prolyl isomerase Pin1 (Marcucci et al., 2011). Pin1 recognizes a conserved phosphorylated Ser/Thr-Pro motif in the N-terminal region of ADAR2, and this interaction is required for nuclear localization of ADAR2. Nevertheless, there are no indications that these factors regulate editing specifically during brain development.

Here, we show that developmentally regulated RNA editing in neurons can be explained by an increasing presence of the ADAR2 enzyme in the nucleus. We show that ADAR2 is imported into the nucleus by importin-α4 and is stabilized by a nuclear interaction with Pin1. These factors increase the accessibility of ADAR2 to the editing substrates during neuronal maturation.

We have previously shown that A-to-I RNA editing increases in both coding and non-coding substrates in the developing mouse brain (Ekdahl et al., 2012; Wahlstedt et al., 2009). To study the mechanisms that regulate A-to-I editing during neuronal maturation, we isolated cortical neural progenitor cells from mouse cortices at embryonic day 16 and cultured them for up to 14 days in vitro (DIV). During this timeframe, neural progenitors matured and formed an extensive neurite network (Fig. 1A; Fig. S1A). To determine if editing increased in a similar way in developing cultured neurons as in the brain, coding as well as non-coding RNA substrates with both specific and overlapping specificity for ADAR1 and ADAR2 editing were selected for analyses after 1–12 days in culture (Ekdahl et al., 2012; Nishimoto et al., 2008; Ohlson et al., 2007). Similar to that in the embryonic brain, low levels of editing were observed at the I/M site in the coding region of the Gabra-3 transcript and at the +6 site in the miRNA precursor transcript pri-miR-376b after 1 DIV (Fig. 1B). As the neurons matured, editing levels gradually increased, and at 12 DIV, the amounts of transcripts edited resembled those detected in the adult mouse brain. Gabra-3 and pri-miR-376b are substrates for both ADAR1 and ADAR2 (Fig. S1B; Ohlson et al., 2007). Editing at the +4 site in the ADAR1-specific non-coding substrate pri-miR-381 also increased as the neurons matured (Fig. 1B; Fig. S1B). Interestingly, editing at the lysine-to-glutamate (K/E) site in the cytoplasmic FMR1 interacting protein 2 (Cyfip2), appeared later in the developing neurons than editing at the other sites – only after 6 DIV a clear G peak was visible in the chromatogram (Fig. 1B). At 12 DIV, 50% of the Cyfip transcripts were edited at the K/E site, which has been shown to be exclusively edited by ADAR2 (Nishimoto et al., 2008; Rueter et al., 1999). To determine if this is a specific pattern for ADAR2-mediated editing, we analyzed auto-editing at the −1 site in the ADAR2 pre-mRNA. Editing at the −1 site gave rise to alternative splicing and a subsequent inclusion of 47 nucleotides in exon 5 (Fig. 1C) (Rueter et al., 1999). Editing-induced alternative splicing was measured by performing semi-quantitative reverse-transcriptase (RT)-PCR, using exonic primers that amplify both normal and alternatively spliced ADAR2 transcripts. Using this approach, we could confirm that editing mediated by ADAR2 gradually increases during neuronal maturation (Fig. 1D). Similar results have been shown previously in primary rat cortical neurons during maturation (Hang et al., 2008). Taken together, these findings show that the editing of several ADAR targets increases during maturation of cultured neurons in a similar manner to that which we have previously shown in the developing mouse brain.

Nuclear RNA editing substrates are edited co-transcriptionally, before splicing, since intronic sequences often are needed to form the double-stranded structure required for editing. The low efficiency of editing during early brain development may therefore be explained by the ADAR enzymes not being located in the same subcellular compartment as the substrate – i.e. not in the nucleus – and/or low expression of the ADAR enzymes in embryos. The ADAR2 subcellular localization has been described as being exclusively nuclear, with an accumulation in the nucleolus, in adult rat neurons and cell lines, such as HeLa and Cos-7 (Desterro et al., 2003; Sansam et al., 2003). Co-immunostaining of immature cortical neurons at 1 DIV showed that ADAR2 localized to the soma and nucleus, with no apparent overlap with the nucleolar fibrillarin protein (Fig. 2A). After 6 DIV, ADAR2 localized predominantly to the nucleus, and pronounced colocalization between fibrillarin and ADAR2 in nucleoli could be detected. After 12 DIV, ADAR2 was almost exclusively localized to the nucleus with an accumulation in the nucleoli in most cells (Fig. 2A). The subcellular localization of ADAR1 was less clear because the two isoforms of the enzyme, p110 and p150, are both cytoplasmic and nuclear in immature as well as in mature neurons. Nevertheless, we observed a clear increase in nuclear localization of ADAR1 in neurons at 12 DIV compared to that in immature neurons at 1 DIV, whereas ADAR1 expression levels remained constant (Figs S2 and S3). To examine the possibility that nuclear levels of ADAR enzymes underlie the regulation of editing during neuronal development, we focused our studies on ADAR2. Western blot analysis of nuclear and cytoplasmic extracts prepared from neurons that had been cultured for 2 and 12 DIV independently confirmed a 2.5-fold increase of ADAR2 protein in the nucleus at 12 DIV compared to that at 2 DIV (Fig. 2B,C). Moreover, the ADAR2 total protein expression level increases after 3 DIV and remains high until 12 DIV (Fig. S2). Taken together, these results indicate a dual effect of both increased nuclear localization of ADAR2, as well as an increased expression as neurons mature.

Selective import of proteins into the nucleus is mediated by specific amino acid sequences known as nuclear localization signals (NLSs). A long stretch of N-terminally positioned residues (residues 4–72) in ADAR2 has been shown, by using deletion mutants, to be crucial for nuclear localization (Desterro et al., 2003; Wong et al., 2003). To refine the sequence that is required for import of ADAR2, we used the first 72 amino acid residues of rat ADAR2 as the input into several NLS in silico prediction programs (Fig. S4A,B) (Kosugi et al., 2009; Nakai and Horton, 1999; Nguyen Ba et al., 2009). The region spanning amino acid residues 49–72 contains two basic clusters spaced by 14 residues, which resemble the motif for either two monopartite or one bipartite NLS (Dingwall and Laskey, 1991; Kalderon et al., 1984; Robbins et al., 1991). Moreover, multiple sequence alignment analysis of the region containing the predicted NLSs showed that the first cluster (52–53) and the basic amino acids of the second cluster (68–72) are conserved between 11 selected mammalian species (Fig. S4C). We tested the function of the predicted NLS clusters by deleting sequences encoding amino acids 48–72 and 4–47 in the context of N-terminally FLAG-tagged ADAR2 (Fig. 3A,B). Subcellular localization was determined by performing immunofluorescent microscopy after transient expression of NLS mutant constructs in human embryonic kidney 293 (HEK293) cells. Expression of full-length fusion proteins was confirmed by western blotting (data not shown). Nuclear localization depended on residues 48–72 but not on residues 4–47 (Fig. 3B). Fusion of rat ADAR2 residues 48–72 to EGFP demonstrated that this sequence was sufficient to target the protein to the nucleus (Fig. 3C). Separately introducing alanine mutations in the first cluster, ⁵²KR⁵³ (ΔNLS1), or in the second cluster, ⁶⁸KKRRK⁷² (ΔNLS2), had only small effects on preventing nuclear localization, whereas combining the mutations (ΔNLS1+2) fully prevented nuclear localization (Fig. 3C). Similarly, the ΔNLS1+2 mutations prevented nuclear accumulation of FLAG-tagged ADAR2 protein (Fig. 3B). We conclude that ADAR2 uses the two highly conserved basic regions in its N-terminal region for its nuclear targeting.

Import of proteins into the nucleus depends on interactions between NLSs and importin-α adaptor proteins. To evaluate which importin-α isoform interacts with ADAR2 in neurons, we focused on two importin-α isoforms that have been previously shown to have affinity towards human ADAR2 in a yeast two-hybrid assay (Maas and Gommans, 2009). By performing co-immunoprecipitation (co-IP) experiments using transiently expressed FLAG-tagged ADAR2 in HEK293 cells, we found that ADAR2 specifically interacts with importin-α4 but not with importin-α5 (Fig. 4A). Importin-α1 was used as a negative control as it has previously been shown to lack binding affinity to ADAR2 (Maas and Gommans, 2009). We further show that the ADAR2–importin-α4 interaction was lost in the nuclear-import-incompetent mutant, Δ48–72, while there was a stronger interaction of importin-α4 with the nuclear-import-competent Δ4–47 mutant (Fig. 4B), possibly due to an altered nuclear import–export rate, increasing the accessibility of the NLS in the cytoplasm. A previous study has demonstrated that importin-α4 mRNA is developmentally upregulated in mouse brain (Hosokawa et al., 2008). To determine if importin-α4 protein expression is regulated during neuronal maturation as described for other importin-α isoforms (α1, α3 and α5) (Yasuhara et al., 2007), we performed western blot analysis on total protein extracts during neuronal maturation. We observed a 60% increase in importin-α4 protein expression over 12 days of neuronal maturation in vitro (Fig. 4C). To confirm an interaction between ADAR2 and importin-α4 in the developing neurons, we used an in situ proximity ligation assay (PLA). In this assay, only proteins within close proximity are detected using rolling circle amplification (Söderberg et al., 2006). Using antibodies against ADAR2 and importin-α4, specific interactions could be visualized as dots by immunofluorescence (Fig. 4D). At 2 DIV, only a few interactions were detected, mostly in the cytoplasm, whereas after 6–9 DIV, an increasing number of interactions were detected in the nucleus. In summary, this result shows that importin-α4 interacts more frequently with ADAR2 as the neuron matures, plausibly bringing it into the nucleus.

To determine if the expression level of importin-α4 influences the level of RNA editing, we transiently co-expressed an editing reporter and an ADAR2 expression vector with and without an importin-α4 expression vector in SH-SY5Y neuroblastoma cells. The editing reporter was a rat Adar2 minigene expressing exon 4, intron 4 and exon 5 (A2-Ex4-5), in which editing leads to alternative splicing as described above (Rueter et al., 1999). Editing was detected by performing RT-PCR using primers amplifying both the normal (unedited) and the alternatively spliced (edited) variants (Fig. 5A). Overexpression of importin-α4 resulted in a 2.2-fold increase in editing of the reporter, indicating that the expression level of importin-α4 affects nuclear editing (Fig. 5B,C). Protein expression of transiently expressed ADAR2 and importin-α4 was assayed using western blot (Fig. 5B). Subcellular localization of the transiently expressed ADAR2 was as expected – localized to the nucleus and the nucleolus (data not shown). Thus, an elevated level of importin-α4 resulted in increased ADAR2-specific editing in neuronal cells, indicating that the level of importin-α4 regulates editing efficiency during neuronal maturation.

Interaction between the proline isomerase Pin1 and phosphorylated ADAR2 stabilizes ADAR2 in the nucleus and has been shown to be required for efficient editing of endogenously and transiently expressed GluA2 transcripts in cell lines (Marcucci et al., 2011). We wanted to investigate the role of Pin1 for the function of ADAR2 in developing neurons and therefore asked if Pin1 expression is altered during maturation. Indeed, western blot analysis showed that Pin1 was expressed at low levels at 1–3 DIV and that the expression increased during maturation, up to 13 DIV (Fig. 6A). Importantly, this was also observed during mouse brain development (Fig. 6B). Furthermore, we determined the subcellular localization of Pin1 in relation to that of ADAR2 during maturation of neurons by immunocytological staining. At 1 DIV, Pin1 was mostly found to localize to the cytoplasm with only a diffuse staining pattern in the nucleus (Fig. 6C). At this stage, ADAR2 co-localized with Pin1 at defined spots in the soma. As the neurons matured, Pin1 intensely stained the cell nucleus but was also present in the cytoplasm. In the mature neurons, Pin1 and ADAR2 colocalization was exclusively seen in the nucleus. Specific interactions between Pin1 and ADAR2 in the developing mouse neurons were analyzed using PLA. At 1 DIV, only a few interactions between ADAR2 and Pin1 were detected, mostly in the cytoplasm (Fig. 6D). At 6 DIV, several interactions were detected in the nucleus, but some complexes were also found in the soma. However, at 13 DIV, a dramatic increase in the number of interactions was observed, mostly in the nucleus. Taken together, the data indicate that nuclear interactions between Pin1 and ADAR2 increase during neuronal maturation.

Several previous analyses have revealed that editing in general increases during neuronal maturation, which has profound effects on both development and functionality of the mammalian brain (Ekdahl et al., 2012; Shtrichman et al., 2012; Wahlstedt et al., 2009). We have previously seen that the protein levels of ADAR1 and ADAR2 are constant during mouse brain development, not following the gradual increase of A-to-I editing (Wahlstedt et al., 2009). In this study, we have addressed the mechanism behind developmentally regulated A-to-I RNA editing. We found low levels of total and nuclear ADAR2 in immature cortical neurons that increase during maturation in accordance with increased nuclear editing. To date, regulation of the ADAR proteins has mainly been studied in non-neuronal cell lines. Although a substantial amount of information can be obtained from analysis in cell lines, it may not reflect dynamic processes within a developing neuron. We therefore analyzed editing and regulation of the ADAR2 protein during maturation of primary neurons and showed that editing of specific targets increases in vitro in a similar way to that in the developing brain. Many ADAR substrates must be edited in their premature nuclear transcript state as the editing reaction requires the presence of intronic sequences to form the necessary double-stranded structures. Thus, cytoplasmic mature mRNAs or miRNAs are no longer substrates for A-to-I editing, and the concentration of nuclear ADAR enzymes is crucial for editing efficiency. In a recent report, it has been shown that sites where editing increases during brain development are more sensitive to fluctuations in ADAR expression than other sites (Hwang et al., 2016). In neurons, we saw a less frequent presence of ADAR2 in the nucleus of immature neurons compared to mature neurons. A reasonable explanation for the low level of editing during early maturation is therefore the low concentration of nuclear ADAR2 protein. Moreover, even though we have chosen to analyze ADAR2, ADAR1-specific editing also increases as neurons mature and, accordingly, a nuclear accumulation of ADAR1 was seen during maturation although the total protein levels remained constant (Figs S2 and S3). This result indicates that the subcellular localization of ADAR1 plays an important role in the regulation of nuclear editing.

A previous study suggests that ADAR2 has a non-canonical NLS within its first 64 amino acids, but this is not enough for nucleolar localization, which requires more than the first 139 amino acids (Desterro et al., 2003). In our analysis, we detected two short conserved NLS sequences in ADAR2 between amino acids 48 and 72. These NLS sequences are functionally redundant since only one of them was enough for nuclear localization of GFP fused to amino acids 48–72. Strikingly, these results explain why a previous deletion of amino acids 64 to 75 in a study by Desterro et al. (2003) retained ADAR2 in the nucleus. This indicates that there are two nuclear localization signals in the N-terminus of ADAR2, but a nucleolar localization signal elsewhere in the protein. Most likely, the dsRNA-binding domains of ADAR2 interact with ribosomal RNA and retain the protein in the nucleolus as ADAR2 is excluded from the nucleolus upon overexpression of dsRNA or through disruption of its RNA-binding capacity (Desterro et al., 2003; Sansam et al., 2003).

In a previous yeast two-hybrid screen, ADAR2 was shown to interact with both importin-α4 and importin-α5 (Maas and Gommans, 2009). In agreement with this, we show that importin-α4 interacts with ADAR2, an interaction that is dependent on the region necessary for nuclear import. However, we could not confirm any binding affinity towards importin-α5, which might be due to differences in the experimental techniques. Furthermore, we saw an increase in interactions between ADAR2 and importin-α4 in developing neurons, as well as an increase in the protein expression of importin-α4. This is in accordance with increased ADAR2-specific editing after elevating importin-α4 expression in neuroblastoma cells. Our result therefore explains how editing can increase during development through increased import of ADAR2, elevating the nuclear concentration.

Intriguingly, ADAR2-specific editing is post-translationally regulated by the phosphorylation-dependent isomerase Pin1 (Marcucci et al., 2011). We show that Pin1 expression is developmentally regulated in cultured primary neurons, which is in line with previous results that have shown a strong induction of Pin1 during neuronal differentiation (Hamdane et al., 2006). Accordingly, we detected an increasing number of interactions between ADAR2 and Pin1 during neuronal maturation. The interaction between Pin1 and ADAR2 depends on phosphorylation of a Ser/Thr-Pro motif in the N-terminal region of ADAR2 (Marcucci et al., 2011). The kinase responsible for this action is currently unknown and possibly, a developmentally timed phosphorylation event induces the conformational change in ADAR2 that triggers the stabilization of ADAR2 in the nucleus. In all, this will contribute to the high level of nuclear ADAR2 in mature neurons. Recently, ADAR-related hypo-editing has been found in Alzheimer's disease (AD) brains with no corresponding decrease in ADAR mRNA levels (Gaisler-Salomon et al., 2014; Khermesh et al., 2016). Also, decreased nuclear levels of Pin1 have been described in brain regions that are known to be prone to neurodegeneration during AD (Liou et al., 2003; Lu et al., 1999). This may be a reflection of dysfunctional Pin1 post-translational control of nuclear ADAR2 concentration in affected neurons in AD. Taken together, our findings explain the restraint on A-to-I editing during early brain development, which occurs as a result of the limited nuclear concentration of the ADAR2 editing enzyme that is due to lack of interactions with importin-α4 and Pin1 (Fig. 7). During development the nuclear import of ADAR2 is mediated by the gradual increase in importin-α4 expression and stabilized by an increased expression of Pin1. As a result, the concentration of ADAR2 in the nucleus is elevated and editing increases.

Our findings suggest a mechanism of ADAR editing regulation that can rapidly change the editing status of multiple substrates simultaneously. This may be important not only for neuronal development but also for neuronal plasticity. Indeed, acute and chronic neuronal activation of both hippocampal and cortical neurons alters the editing of several substrates (Balik et al., 2013; Sanjana et al., 2012). However, differences in nuclear concentration of ADAR2 upon neuronal activation remain to be investigated.

Primary neurons were prepared from cortices dissected from embryonic-day-16 pregnant NMRI mice (Charles River). After trituration and filtration through a 40 μm membrane, cells were plated on poly-D-lysine- (Sigma) coated vessels and grown in Neurobasal^® medium supplemented with 2% B27, 0.5 mM L-glutamine, 20 U/ml penicillin-streptomycin (Invitrogen) and 12.5 μM L-glutamate (Sigma). All animal procedures were approved by the Animal Ethics Committee of the North Stockholm region. HEK293, SH-SY5Y and N2a cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and 100 U/ml penicillin-streptomycin (Invitrogen). Cells were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.

Total RNA was isolated from cell cultures using the mammalian total RNA kit (Sigma) or Trizol (Invitrogen). After DNase I (Sigma) treatment, the RNA was reverse-transcribed using SuperScript II and random hexamers (Invitrogen). For editing analysis of endogenous transcripts from neuronal cells, RT-PCR was performed using gene-specific primers and Phusion DNA polymerase (ThermoScientific). Editing frequencies were determined by Sanger sequencing (Eurofins MWG Operon) of gel purified RT-PCR amplicons. Analysis of editing in the editing-splicing reporter A2-Ex4-5, cDNA synthesis, DNase I treatment and RT-PCR using Phusion DNA polymerase (ThermoScientific) was performed as above. The fold-change in the amount of editing-induced alternative over that of the normally spliced transcript was assayed by measuring the intensities of the PCR products using ImageLab (Bio-Rad).

Pri-miR-381 and pri-miR-376b constructs were constructed from PCR-amplified mouse genomic DNA using primers approximately 200 nucleotides upstream and downstream of the annotated pre-miRNA. Amplicons were cloned into pcDNA3. All point mutations and deletion constructs were generated through QuickChange site-directed mutagenesis (Stratagene) according to the manufacturer's instructions using the rat pcDNA3-FLAG-ADAR2 as template (Bratt and Öhman, 2003). To construct GFP-rADAR2-NLS, the rat ADAR2 residues 48–72 were subcloned into the EcoRI and SalI sites of the enhanced GFP (pEGFP-C2, CLONTECH) vector. The editing-splicing reporter pRC-CMV-A2 exon 4–5 (A2-Ex4-5) plasmid has been described previously (Rueter et al., 1999). The pCMVTNT-T7-KPNA3 plasmid was obtained from Addgene (plasmid # 26679) and is originally described in Kelley et al. (2010). The pcDNA3-Pin1-HA plasmid was a gift from Mary O'Connell (Masaryk University, Brno, Czech Republic).

Cells grown on glass slides for the indicated times were fixed in 4% PFA and permeabilized with 0.05% SDS. After blocking with 3% BSA, slides were incubated with the following primary antibodies overnight at 4°C: mouse monoclonal anti-βIII-tubulin (1:1000, SDL.3D10, Sigma), rabbit polyclonal anti-ADAR2 (1:200, Sigma), mouse monoclonal anti-ADAR1 (1:500, 15.8.6, Santa Cruz Biotechnology), mouse monoclonal anti-fibrillarin (1:200, 38F3, GeneTex), mouse monoclonal anti-Pin1 (1:200: G-8, Santa Cruz Biotechnology), mouse monoclonal anti-KPNA3 (1:200, B-1, Santa Cruz Biotechnology), rabbit anti-FLAG (1:3000, Sigma). For immunofluorescence, cells were incubated with anti-mouse-IgG conjugated to Alexa-Fluor-488 and/or anti-rabbit-IgG conjugated to Alexa-Fluor-555 (Invitrogen) and mounted using Prolong Gold antifade reagent containing DAPI (Invitrogen). In situ Proximity Ligation Assay (Duolink^®) was conducted following the manufacturer's instructions (Sigma). Images were acquired using an Axiovert 200M inverted fluorescence microscope (Carl Zeiss) equipped with a Plan-Apochromate 63×1.4 oil objective lens.

Nuclear and cytoplasmic extracts were prepared using NE-PER nuclear and cytoplasmic extraction kit (Thermo Scientific) according to the manufacturer's instructions. For total protein extracts, cells were lysed using Lysis-M reagent (Roche) supplemented with protease and phosphatase inhibitor cocktails (Roche). Total protein concentration was determined with Bradford assay, and samples were separated on 4–15% SDS-polyacrylamide gels (BioRad) and blotted onto PVDF membranes. The blots were probed with the following primary antibodies overnight at 4°C: rabbit polyclonal anti-ADAR2 (1:200, Sigma), rabbit polyclonal anti-ADAR1 (1:500, Sigma), mouse monoclonal anti-Pin1 (1:1000: G-8, Santa Cruz Biotechnology), mouse monoclonal anti-βIII-Tubulin (1:2000, SDL.3D10, Sigma), mouse monoclonal anti-U1-70K (1:1000, H111, Synaptic Systems), mouse monoclonal anti-MAP2 (1:250, HM-2, Sigma), mouse monoclonal anti-KPNA1 (1:200, 187.1, Santa Cruz Biotechnology), mouse monoclonal anti-KPNA2 (1:200, B-9, Santa Cruz Biotechnology), mouse monoclonal anti-KPNA3 (1:200, B-1, Santa Cruz Biotechnology), mouse monoclonal anti-GAPDH (1:10,000, 6C5, Abcam), mouse monoclonal anti-nestin (1:500, rat-401, Millipore), guinea-pig polyclonal anti-NeuN (1:500, Millipore), mouse monoclonal anti-PSD-95 (1:500, 7E3-1B8, Millipore), rat monoclonal anti-GFAP (1:500, 2.2B10, Invitrogen), mouse monoclonal β-actin (1:10,000, AC-15 Sigma). Horseradish peroxidase (HRP)-conjugated secondary antibodies (DAKO) were used at a 1:2000 dilution. Chemiluminescent detection of protein was performed using the WesternBright Sirius detection system (Advansta), and band intensities were quantified using ImageLab (BioRad). For total protein measurements, a sister polyacrylamide gel was stained with Coomassie Blue (R-250) and destained according to standard procedures. To quantify nuclear expression or total protein content, intensities of bands or lanes, were measured using Bio-Rad Image Analysis. For co-immunoprecipitation, cells were resuspended in lysis buffer [125 mM NaCl, 50 mM Tris-HCl pH 7.5, 1% NP-40, 1 mM EGTA and 10% glycerol, and supplemented with Complete protease and phosphatase inhibitor cocktail (Roche)]. After removal of insoluble material, the supernatants were incubated with rabbit anti-FLAG (1:2000, Sigma) followed by addition of Sepharose-G-breads (Invitrogen).

All numerical data are presented as means±s.e.m., calculated using Prism 7 software (GraphPad). Statistical significance was assessed using paired two-tailed Student's t-test or one-way ANOVA, indicated in the respective figure legends. P-values less than 0.05 were considered statistically significant.

We thank Mary O'Connell for invaluable scientific discussions, Anna-Stina Höglund for microscopy technical support, Roger Karlsson (Stockholm University, Stockholm, Sweden) for kindly donating the pEGFP plasmid and Claes Andréasson for critically reading the manuscript. We acknowledge the Imaging Facility at Stockholm University (IFSU) for support with microscopy.