Subcellular Distribution of ADAR1 Isoforms Is Synergistically Determined by Three Nuclear Discrimination Signals and a Regulatory Motif*

ADAR1 is an RNA-specific adenosine deaminase that edits RNA sequences. We have demonstrated previously that different ADAR1 isoforms are induced during acute inflammation. Here we show that the mouse ADAR1 isoforms are differentially localized in cellular compartments and that their localization is controlled by several independent signals. Nuclear import of the full-length ADAR1 is predominantly regulated by a nuclear localization signal at the C terminus (NLS-c), which consists of a bipartite basic amino acid motif plus the last 39 residues of ADAR1. Deletion of the NLS-c causes the truncated ADAR1 protein to be retained in the cytoplasm. The addition of this sequence to pyruvate kinase causes the cytoplasmic protein to be localized within the nucleus. The localization of nuclear ADAR1 is determined by a dynamic balance between the nucleolar binding activity of the nucleolar localization signal (NoLS) in the middle of the protein and the exporting activity of the nuclear exporter signal (NES) near the N terminus. The NoLS consists of a typical monopartite cluster of basic residues followed by the third double-stranded RNA-binding domain. These signals act independently; however, NES function can be completely silenced by the NLS-c when a regulatory motif within the catalytic domain and the NoLS are deleted. Thus, the intracellular distribution of the various ADAR1 isoforms is determined by NLS-c, NES, NoLS, and a regulatory motif.

ADAR1 is an RNA-specific adenosine deaminase that edits RNA sequences. We have demonstrated previously that different ADAR1 isoforms are induced during acute inflammation. Here we show that the mouse ADAR1 isoforms are differentially localized in cellular compartments and that their localization is controlled by several independent signals. Nuclear import of the full-length ADAR1 is predominantly regulated by a nuclear localization signal at the C terminus (NLS-c), which consists of a bipartite basic amino acid motif plus the last 39 residues of ADAR1. Deletion of the NLS-c causes the truncated ADAR1 protein to be retained in the cytoplasm. The addition of this sequence to pyruvate kinase causes the cytoplasmic protein to be localized within the nucleus. The localization of nuclear ADAR1 is determined by a dynamic balance between the nucleolar binding activity of the nucleolar localization signal (NoLS) in the middle of the protein and the exporting activity of the nuclear exporter signal (NES) near the N terminus. The NoLS consists of a typical monopartite cluster of basic residues followed by the third doublestranded RNA-binding domain. These signals act independently; however, NES function can be completely silenced by the NLS-c when a regulatory motif within the catalytic domain and the NoLS are deleted. Thus, the intracellular distribution of the various ADAR1 isoforms is determined by NLS-c, NES, NoLS, and a regulatory motif.
ADAR1 deaminates adenosines in mRNA, thereby altering codons and giving rise to different protein isoforms (1,2). This deaminase activity was first identified in Xenopus embryos, which unwind dsRNA 1 by converting adenosines to inosines (3,4). A family of A-to-I RNA editing enzymes exists in mammals (5)(6)(7)(8)(9) with homologues in Drosophila (10), zebrafish (11), and Xenopus (12), suggesting that these enzymes are evolutionarily conserved. ADAR1 is ubiquitously expressed in a variety of cells and tissues (6,8) and may participate in host defense by targeting viral RNA (13). It was recently shown that ADAR1 participates in the development of acute inflammation (14,15). ADAR1 knock-out mice die as embryos with immature erythrocytes (16), suggesting that ADAR1-mediated RNA editing is critical for normal proliferation and/or differentiation of erythrocytes during development.
The sequences of ADAR1 from human, rat, and mouse are highly conserved; a Z-DNA-binding motif, three dsRNA-binding domains (dsRBDs), and a conserved deaminase domain have been identified. Full-length human ADAR1 protein is found in the cytoplasm; it is initially transported into the nucleus and then re-exported by a nuclear export signal (NES) near the N-terminal region (17). An atypical nuclear localization signal (NLS) with nuclear import activity occurs within dsRBDIII so that human ADAR1 displays the characteristics of a shuttling protein (18,19). A different NLS within the Nterminal 269 amino acids is reported to direct active import of human ADAR1 (17). In addition, a human ADAR1 fragment lacking the NES has been shown to localize predominantly within the nucleus (19). A similar fragment of human ADAR1 was found exclusively in the nucleus, implicating another NLS at the C-terminal end of human ADAR1 (17). The existence of a nucleolar localization signal (NoLS) further complicates the localization of ADAR1. The NoLS is capable of targeting ADAR1 to the nucleolus where rRNA interacts with imported ribosomal proteins to form preribosomal particles (20). Nucleolar localization of ADARs may suggest that RNA editing occurs during transcription or RNA processing in the nucleolus. Alternatively, it may regulate protein synthesis (21). Furthermore, the Xenopus ADAR1 contains a distinctive NLS downstream from the Z-DNA-binding motif, which targets the enzyme to the nascent ribonucleoprotein matrix on lampbrush chromosomes, where it is specifically associated with active transcriptional sites (22,23). This observation suggests that RNA editing by Xenopus ADAR1 is coupled with transcriptional events or targets newly synthesized RNAs. Since human ADAR1 contains an NES that has not been identified in Xenopus ADAR1, this protein may function differently in mammalian cells. Nevertheless, it is likely that RNA editing by ADAR1 is dynamically regulated in different cellular compartments by these localization signals.
Certain patterns of basic amino acid residues are necessary for proteins to be imported into the nucleus, although no precise consensus amino acid sequence has been identified (24). Many variable NLS sequences have been identified in viral and cellular proteins; these sequences are classified as monopartite or bipartite (25,26). The typical monopartite NLS is a cluster of basic residues starting with Pro and followed by 5 residues, of which at least 3 are either Lys or Arg. The bipartite pattern (e.g. KRXKKXXKX) begins with 2 basic residues (Lys or Arg) followed by a 10-residue spacer and finally a cluster in which at least 3 out of 5 residues are Lys or Arg (25). Proteins containing these signals are transported into the nucleus by ATP-dependent translocation through the nuclear pore complexes (27, 28). A number of NLS-binding proteins have been reported to function as cytoplasmic receptors that deliver karyophilic proteins to the transport machinery of nuclear pore complexes. No precise NoLS consensus amino acid sequence has been identified, although NoLS signals usually reside within an NLS sequence. An atypical NoLS in human immunodeficiency virus type 1 contains two stretches of basic amino acids (GRKKRRQR-RRAHQN, basic residues are underlined) to target the Tat protein to the nucleolus (29).
We have shown previously that multiple short versions of ADAR1 are induced during acute inflammation (30). Deletions often occur in these inducible isoforms within the regions that are important for nuclear export or import, suggesting that the intracellular localization of ADAR1 may be differentially regulated under various pathological conditions. Here we show that localization of the ADAR1 isoforms is regulated by several nuclear discrimination signals. Furthermore, nuclear import of the full-length ADAR1 is predominantly determined by an essential NLS within the last 56 residues of the C terminus. A regulatory motif may be required for nuclear export of the NES, and this motif can be completely silenced by the NLS. Finally, we show that a signal near the dsRBDIII mediates nucleolar recruitment of the short version of ADAR1.

MATERIALS AND METHODS
Western Blotting-Samples containing 60 g of total protein derived from thymocytes of BALB/c mice (Jackson Laboratories) were resolved on 10% SDS-PAGE and transferred to a polyvinylidene difluoride membrane. ADAR1 immunoreactivity was detected using rabbit antiserum raised against the C terminus of recombinant mouse ADAR1 expressed in Escherichia coli (2765-3459, AF291050) or against the N terminus of a synthetic mouse ADAR1 peptide (Santa Cruz Biotechnology).
Construction of ADAR1 Chimeras-Different mADAR1 fragments were generated by PCR using the pCRII-ADAR1Lb plasmid (AF291050) as a template. A BamHI sequence was added to each end of the amplified fragments. The 5Ј-primers included a start codon with a KoZak consensus sequence ((A/G)NNATGG) (31). The stop codon in the 3Јprimer was removed to create an in-frame fusion to enhanced green fluorescent protein (EGFP). The primer sequences are shown in Table I. PCR products were then digested with BamHI and directly cloned into pEGFP-N1 or pEGFP-N2 vectors (Clontech) at the N-terminal end of EGFP. For the ⌬NLS and NoLS-I constructs, fragments were excised from the mADAR1 cDNA with the following restriction enzymes: HindIII or XbaI/HindIII, respectively. The resulting inserts were then ligated into the pEGFP-N1 or pEGFP-N2 vector. Positive clones were identified by BamHI digestion, and the reading frames were confirmed by sequencing. ADAR1-EGFP DNAs with the correct sequences were transfected into mouse fibroblasts (3T3) or other cell lines, and the expression of protein chimeras was confirmed by the presence of green signals under fluorescence or confocal microscopy.
Two chimeric pyruvate kinase constructs, Myc-PK-NLS and Myc-PK-NoLS-NLS, were generated to test the function of NLS and NoLS. The NLS fragment, which consists of the last 56 residues of ADAR1, was amplified by PCR to add EcoRI and XhoI sites at its 5Ј-and 3Ј-ends, respectively. This fragment was then ligated into the EcoRI and XhoI sites of Myc-PK pcDNA3 (a kind gift from Dr. M. Matunis, The Johns Hopkins University) (32). The NoLS fragment (including the P-cluster and dsRBDIII) was amplified to add an EcoRI site at both ends; this fragment was ligated into the EcoRI site of Myc-PK-NLS pcDNA3. These constructs were then transfected into 3T3 cells, and localization of the chimeric proteins was evaluated by immunofluorescence with an anti-Myc antibody.
Construction of Bifunctional Chimeras-The NES fragment (385-603, GenBank TM accession number AF291050), which contains the leucine-rich nuclear export consensus (cNES) consensus as well as the N-terminal bipartite signal (Bipartite N) and the Z-DNA-binding domain (Z␣), was amplified by PCR using the pCRII-ADAR1Lb plasmid as a template. XhoI and BamHI sites were added to the primers to generate restriction enzyme sites at the 5Ј-or 3Ј-ends, respectively. The amplified fragment was then subcloned into the BamHI site of the pEGFP-N1 vector to generate the pEGFP-NES construct. Various NLS-c fragments, extending from 3229, 2983, 2741, 2414, or 1975 to the stop codon of mouse ADAR1, were amplified by PCR using the same template. The stop codon was replaced with TGG to generate chimeras in-frame with EGFP. A BamHI was added to each end of the amplified fragment, and the insert was then ligated into the unique BamHI site of the pEFGP-NES construct to generate pEGFP-NES-NLS bifunctional chimeras. Similarly, the NoLS fragments were amplified from 1983 to 2414 or 1983 to 2263 of mouse ADAR1 and inserted into the pEFGP-NES construct to generate pEGFP-NES-NoLS bifunctional chimeras. After transformation and purification, clones with inserts were selected by restriction enzyme digestion, and the reading frames were confirmed by DNA sequencing.
Cell Culture and Transfection-Mouse fibroblasts (3T3), mouse neuroblastoma cells (N18), human HeLa, or embryonic kidney 293 cells were cultured to the logarithmic stage in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Approximately 2 ϫ 10 5 cells were placed in each well of 24-well plates and cultured overnight in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. In a typical transfection, 2 g of the pEGFP-ADAR1 chimeric construct was mixed with 3 l of LipofectAMINE 2000 (Invitrogen) in 0.5 ml of serum-free Opti-MEM, according to the manufacturer's recommended protocol. After incubation at 37°C for 5 h, the cells were washed and recultured with fresh Dulbecco's modified Eagle's medium containing 10% fetal bovine serum for 12-24 h. Green fluorescence usually appeared 7-9 h after transfection; photographs were taken 12-24 h after transfection using a fluorescence or confocal microscope (Zeiss).
Blocking CRM1-mediated Nuclear Export by Leptomycin B-To block nuclear export with leptomycin B (LMB, a kind gift of Minoru Yoshida, University of Tokyo), 3T3, N18, or 293 cells (2 ϫ 10 5 cells/well) were incubated for 5-6 h with 2 g of pEGFP-ADAR1 DNA and 6 l of LipofectAMINE 2000 (Invitrogen), as described above. After washing, the transfected cells were cultured in appropriate medium in either the presence or the absence of LMB (1 ng/ml). Green fluorescence was examined 12-24 h after transfection.

ADAR1 Variants Truncated at the N and C Termini
Are Differentially Localized-We have shown that two major short forms of mouse ADAR1 truncated from the N-terminal (mADAR1 p100 and p80) are expressed in spleen and thymus during endotoxin-induced acute inflammation (30). Using an antibody directed against the N-terminal region of ADAR1, we also detected short isoforms truncated from the C-terminal (Fig. 1). The anti-C-terminal antibody detected full-length ADAR1 and isoforms truncated from the N terminus; the anti-N-terminal antibody detected full-length ADAR1 and two isoforms truncated from the C terminus. We reasoned that the isoforms derived from truncation of the N and C termini might be localized in different cellular compartments. mADAR1 chimeras containing fragments truncated from the N or C terminus of mADAR1 and fused to EGFP (Fig. 2, ⌬NES and ⌬NLS) were transiently expressed in various cell lines, and their localization was determined by fluorescence microscopy. Consistent with previous studies, full-length mADAR1 was found in the cytoplasm ( Fig. 2A) (18,30), whereas the ADAR1 fragment from which the N-terminal region was deleted was localized in the cytoplasm and nucleolus (Fig. 2B, ⌬NES) (17,30,33). LMB, an antagonist of the nuclear receptor exportin CRM1 (34 -36), at a concentration of 1 ng/ml blocked export and resulted in accumulation of full-length ADAR1 within the nucleus ( Fig. 2A, ϩLMB). The fragment lacking the C-terminal region was found within the cytoplasm (Fig. 2C, ⌬NLS); however, LMB did not cause ⌬NLS to accumulate in the nucleus, even at concentrations up to 10 ng/ml (Fig. 2C, ϩLMB). The ADAR1 variant ⌬NLS, in which the C terminus was truncated, was retained in the cytoplasm after translation without nuclear importation; deletion of the last 56 amino acid residues prevented mADAR1 from being transported into the nucleus.
Nuclear Discrimination Signals Are Conserved in Mammalian ADAR1-Since NLS activity was also observed near the third dsRBD of human ADAR1 (18,19), we reasoned that multiple signals might exist within mADAR1 that function synergistically in regulating nuclear importation. Using Reinhardt's method (PSORT, bioweb.pasteur.fr/seqanal/interfaces/ psort2), we identified three putative nuclear discrimination signals that are conserved in the human, rat, and mouse ADAR1 sequences (Fig. 3). Two of these are located in the Nand C-terminal regions, with the sequence KRDINRILYSLE-KKGKL located at position 171 and the sequence RRDLLQL-SYGEAKKAAR at position 1097, respectively. Both follow the bipartite pattern (25). The N-terminal bipartite signal (Bipar-tite N) is located within the Z-DNA-binding domain (Z␣) and is adjacent to a cNES; this region was shown to mediate nuclear export of ADAR1 (17)(18)(19). The C-terminal bipartite signal (Bipartite C) is found in the adenosine deaminase domain, within the last 56 residues of ADAR1. A similar region was shown to be important for nuclear translation (21), which we show to be involved in nuclear import. The third signal (P-cluster) is a typical monopartite cluster of basic residues, with a conserved pattern starting with Pro, followed by Lys or Arg in 3 out of 5 residues (PNKIRRI) (37). The P-cluster is adjacent to the third dsRNA-binding domain (dsRBDIII), which has been shown previously to be necessary for nuclear import of human ADAR1 (18,19,33), and which we show in this study is also important for nucleolar localization. These consensus sequences are not found in ADAR sequences from Xenopus, Drosophila, Caenorhabditis elegans, or fishes, indicating that the intracellular lo- FIG. 1. Expression of ADAR1 isoforms truncated at the C terminus and N terminus. Cell extracts were prepared from mouse spleen and were then subjected to Western blotting using antibodies directed against the C-terminal end of ADAR1 (C-term) or the Nterminal end of ADAR1 (N-term). Only the full-length ADAR1 (p150) was detected by both antibodies. The ADAR1 isoforms truncated at the N terminus include p110 and p80, from which the NES region is removed. The isoforms truncated at the C terminus include p100 and p90, from which the NLS-c region is deleted.
FIG. 2. Three ADAR1 isoforms are differentially localized. The cDNAs encoding three ADAR1 isoforms were cloned upstream of EGFP, and the constructs were transfected into 3T3 cells in the absence (left panels) or presence (right panels) of LMB. Localization of transiently expressed chimeras was visualized under a fluorescence microscope (Olympus IMT-2, 488 nm); photographs were taken 12 h after transfection. A, ADAR1, the full-length ADAR1. B, ⌬NES, ADAR1 fragment from which the NES region was deleted; this is also a naturally expressed short ADAR1 isoform (GenBank TM accession number AF291877). C, ⌬NLS-c, ADAR1 fragment from which the NLS region was deleted. Bar, 10 m. calization of ADAR1 may be uniquely regulated in mammals, as compared with other taxa.
The Last 56 Residues Containing the Bipartite C Domain Predominantly Control Nuclear Importation of Mouse ADAR1-We examined the function of Bipartite C by constructing N-terminal deletions of mADAR1. In agreement with our previous findings, fragments without Bipartite N were consistently transported into the nucleus and bound to the nucleolus (Fig. 4A); fragments lacking Bipartite N and all dsRNA-binding domains were still found in the nucleus at large (Fig. 4B). This suggests that Bipartite C is functionally sufficient for nuclear accumulation of ADAR1. To identify the essential residues for nuclear localization, a fragment containing Bipartite C followed by 39 residues at the C-terminal end of ADAR1 was linked to EGFP. In the same cells, this chimera was found in the nucleus (Fig. 4C). However, the nucleolar localization was not observed when Bipartite C and its N-terminal sequences were completely removed (Fig. 4D). Thus, the sequence containing Bipartite C followed by 39 residues at the C-terminal end of ADAR1, termed NLS-c, is an independent and fully functional signal for nuclear localization.
To ascertain whether the nuclear localization activity of NLS-c was mediated by nuclear import rather than nuclear retention, we fused the NLS-c fragment to pyruvate kinase (PK), a well characterized cytoplasmic protein (19,32). As shown in Fig. 5, PK-NLS-c-Myc was localized within the nucleus, whereas the PK-Myc control was detected in the cytoplasm.
The NLS-c Interacts with NES-Since the final localization of ADAR1 may be dynamically balanced by nuclear exportation and importation, we constructed a series of bifunctional ADAR chimeras containing both the NLS-c and NES sequences. An NES fragment containing cNES and Bipartite N (385-603 bp) was linked to a C-terminal fragment containing the minimum NLS-c and followed by GFP. The NES fragment consistently formed a fully functional CRM1-mediated exporter that was exclusively localized in the cytoplasm and accumulated in the nucleus and nucleolus when nuclear exportation was blocked by LMB (Fig. 6, NES). However, the chimeras containing both NES and NLS-c were found within the nucleus (Fig. 6, EN3229  and EN2983). The addition of LMB did not affect the nuclear localization of the EN3229 and EN2983 constructs. Thus, the NLS-c completely masked the function of CRM1-mediated nuclear exportation through an as yet unknown mechanism. Nevertheless, we demonstrated that nuclear exportation of the NES fragment was restored when the NLS-c was further extended to the catalytic domain (Fig. 6, EN2741 and EN2413). These data suggest that a regulatory motif within the catalytic domain (804 -994) is involved in the interaction between the NES and the NLS-c. Since the fragment EN2741 lacking the leucine-zipper-like dimerization domain (19) was still predominantly exported to the cytoplasm or accumulated in the nucleus when LMB was added, the regulatory motif is more likely localized between the residues from 914 to 994.  3. Nuclear discrimination signals in ADAR1. The PSORT program domain similarity search was used to identify putative nuclear discrimination signals in mouse (AF291050), rat (U18942), and human (U75503) ADAR1. The Bipartite N and Bipartite C motifs consist of 2 consensus basic residue clusters separated by 10 residues (25). The P-cluster is a typical monopartite basic cluster (37). cNES matches the consensus sequence of a nuclear export signal (17). The P-cluster and dsRBDIII Provide a Fully Functional Nucleolar Localization Signal-The NLS-c was further extended to include the P-cluster and the dsRBDIII in NES-NLS-c constructs (Fig. 6, EN1975). The resulting chimera was still found within the cytoplasm; however, it was recruited to the nucleolus when LMB was added. Thus, the EN1975 construct shares all known properties for intracellular localization with fulllength ADAR1. This suggests that the NES, NLS-c, NoLS, and the putative regulatory signal represent the elements that are essential for ADAR1 localization. These findings also suggest that the P-cluster followed by the dsRDBIII region could function as an NoLS. When NLS-c was deleted, the chimera containing NES and NoLS remained in the cytoplasm; however, it was recruited into the nucleolus in the presence of LMB (Fig. 6,  EO2414 and EO2263). These data indicate that NoLS containing the P-cluster and dsRBDIII can be transported into the nucleus and then recruited to the nucleolus. Although the nuclear localization activity of the NoLS is consistent with previous observations in human ADAR1 (18,19), it should be noted that this fragment may also diffuse into the nucleus because of its small size. Because the larger fragment containing the NES and NoLS was retained in the cytoplasm even in the presence of LMB (Fig. 2C, CϩLMB), the NLS-c is likely to be the active signal that determines nuclear importation of full-length ADAR1.
Analysis of other constructs further supported the hypothesis that the residues spanning dsRBDIII are important for nucleolar localization. First, the short mADAR1 isoform without Bipartite N was recruited into the nucleolus even in the absence of LMB (Fig. 2B). We also demonstrated that deletions from the N-terminal end of ADAR1 did not alter the nucleolar localization unless the P-cluster and the dsRBDIII were removed; fragments in which sequences beyond the P-cluster and the third dsRBD were deleted were localized within the nucleus (Fig. 7, A-C). In parallel, truncations from the C terminus of ADAR1 abolished nucleolar interactions when the deletions reached the P-cluster and dsRBDIII (Fig. 7, D and E). The requirement of the P-cluster and dsRBDIII for nucleolar localization of mADAR1 was further confirmed by constructing a fragment containing only these sequences (Fig. 7F, NoLS); this construct was efficiently recruited into the nucleolus. In comparison, four P-cluster sequences were found in human ADAR2 and deletion of the second one (PTKKKAK) within 75-132 residues abolished nucleolar binding (33), supporting the idea that this motif is required for the nucleolar localization of ADAR family proteins.
Finally, we also tested whether the identified NoLS was able to recruit a typical cytoplasmic protein to the nucleolus. To this end, Myc-tagged pyruvate kinase was first fused to the NLS-c and then to the NoLS. The chimera was not only imported into the nucleus but also condensed within the nucleolus (Fig. 5C), suggesting that the NoLS acts as an independent nucleolar localization signal.

DISCUSSION
The Last 56 Residues at the C Terminus of mADAR1 Dominate Nuclear Importation of Full-length ADAR1-In this study, we identified a new NLS that includes the last 56 residues at the C-terminal end of ADAR1 and consists of the Bipartite C consensus sequence and 39 adjacent residues. This NLS signal (NLS-c) is essential for nuclear import because mADAR1 variants without this signal cannot be transported into the nucleus, even when CRM1-mediated nuclear export is blocked. This signal is sufficient to transport pyruvate kinase, a typical cytoplasmic protein, into the nuclei. It is therefore both necessary and sufficient for nuclear localization of mADAR1 protein.
Nuclear localization activity has been observed previously in different sites of human ADAR1 (17)(18)(19). Nuclear accumulation of the EGFP-tagged NES fragment in LMB-treated cells sug- FIG. 5. The NLS-c causes pyruvate kinase, which is normally a cytoplasmic protein, to be localized within the nucleus. Constructs encoding Myc-tagged PK (A), Myc-tagged PK followed by NLS-c (B), and Myc-tagged PK (C) followed by NoLS/NLS-c were generated and transfected into HeLa cells. Cells were fixed on glass slides and stained with an anti-Myc antibody. Localization of the chimeras was analyzed by immunofluorescence. NLS-c, Bipartitite C followed by the last 39 residues of ADAR1; NoLS, P-cluster followed by the dsRBDIII .   FIG. 6. Interactions between the NES, NLS-c, and a regulatory motif. Bifunctional chimeras containing the NES and various truncations of the NLS-c were constructed and fused upstream of EGFP. Localization of transiently expressed chimeras was examined in 3T3 cells. Ratios of EGFP distribution in the cytoplasm (cyto), nucleoplasm (Nu), and nucleolus (Nol) were estimated. The percentage of cells with a particular distribution pattern is shown in parentheses. SF, the short form of ADAR1; NoLS, the P-cluster plus the dsRBDIII; NLS-c, the Bipartite C plus 39 downstream residues; NES, the cNES consensus plus the bipartite N and the Z␣; EN3229, EN2983, EN2741, EN2413, and EN1975, the NES followed by the NLS-c fragment truncated at 3229, 2983, 2741, 2413, or 1975, respectively; EO2414 and EO2263, the NES followed by the NoLS fragment truncated from 1893 to 2414 or 2263, respectively; R, the putative regulatory motif.
gested that an NLS resides within the N-terminal region of human ADAR1 (17). Using the PK-Myc reporter system, the dsRBDIII was shown to lead the cytoplasmic protein to the nucleus (18,19). In our study, we also observed nuclear localization activity in fragments containing the NES (Fig. 6, NES) and the dsRBDIII (Fig. 6, EO2414 and EO2263). However, the nuclear localization activity of the NES fragment was observed only when export was blocked by LMB. Removal of the NES did not affect nuclear import of the short forms of ADAR1. We also show that the dsRBDIII, together with the P-cluster, form a fully functional NoLS that mediates the nucleolar accumulation of the short mADAR1; removal of NoLS abolished nucleolar binding of the ADAR1-EGFP chimeras. Our data support the hypothesis that the NoLS in mouse ADAR1 has both nuclear and nucleolar activities, which is also consistent with the fact that NoLS is usually found within NLS and is associated with nuclear localization activity. Among NLS-c, NoLS and NES, our data suggest that NLS-c predominantly controls nuclear importation of full-length ADAR1. This is supported by the finding that ADAR1 isoforms lacking the NLS-c were retained in the cytoplasm even in the presence of LMB. In contrast, ADAR1 fragments from which the NES or the NoLS was deleted could still be found within the nucleus. However, the isoforms truncated at the C terminus, which lack the NLS-c as well as the residues 914 -994, are likely to be imported through the nuclear localization activity of NoLS (18) (Fig. 6, EO2414 and EO2263) and/or NES (17). Therefore, we suggest that the residues spanning 914 -994 represent a putative regulatory motif that interacts with adjacent localization signals. In this study, we have demonstrated that the NLS-c completely masks CRM1-mediated nuclear exporting of the NES when these two signals are closely linked. The regulatory motif has been shown to silence the masking effect of NLS-c through interactions that have not yet been determined. This regulatory motif closely resembles a leucine-zipper-like dimerization domain that was found to interfere with the nuclear accumulation of a fragment near the dsRBDIII of human ADAR1 (19). Since dimerization of ADAR1 is required for editing activity (38,39), the regulatory motif spanning 914 -994 and the leucine-zipper-like dimerization domain near the approximate region of 785ϳ885 may act synergistically on the nuclear localization activity of NLS-c, NoLS, and/or NES in ADAR1. This would be expected to affect RNA editing in various cellular compartments. Further studies will be needed to address these issues.
The P-cluster Followed by the dsRBDIII Is Sufficient for Nucleolar Recruitment-We have identified an independent NoLS in the middle of ADAR1 that consists of a typical monopartite basic residue motif (P-cluster) followed by dsRBDIII. Although tandem repeats of the P-cluster consensus are able to guide proteins to the nucleolus (data not shown), efficient function requires the adjacent dsRBDIII. This NoLS was also able to cause the cytoplasmic protein pyruvate kinase to be imported into the nucleus and condensed within the nucleolus when it was linked to the mADAR1 NLS and NoLS. In addition, the NES fragment (Fig. 6, NES) can be partially recruited to the nucleolus. However, this occurred only when the CRM1 pathway was artificially blocked by LMB treatment.
The same signal in human ADAR1 was found to be important for nuclear localization. We now show that this NoLS is also important for regulating the nucleolar localization of mADAR1 because 1) short ADAR1 isoforms are found predominantly in the nucleolus and 2) the full-length ADAR1 is found in the nucleolus when exportin CRM1 is blocked. Nucleolar localization of endogenous ADAR1 has been detected in human U-cells (40). Furthermore, nucleolar targeting is also observed in human ADAR1 and ADAR2 (33). In human ADAR2, nucleolar targeting requires 55 residues in the first dsRBD (dsRBDI) including two P-clusters: PGRKRP at position 49 and PT-KKKAK at position 125. The second P-cluster at 125 has been shown to be necessary but not sufficient for nucleolar targeting. In contrast, the P-cluster (PNKIRRI) in mouse ADAR1, together with the third dsRBD, represents a fully functional NoLS. The differences in the properties of NoLS between human and mouse ADAR1 may be due, at least in part, to the sequence variation in the P-cluster. Further investigation is needed to determine whether the primary structure or inter- FIG. 7. The NoLS consists of the P-cluster and dsRBDIII. The localization of each transiently expressed EGFP chimera was examined in 3T3 cells. The P-cluster is shown in boldface; the consensus sequence is underlined. The sequence of dsRBDIII is shown in italics. A, ⌬NES-II, a fragment with a deletion from the N-terminal ends to dsRBDI. B, ⌬NES-I, a fragment with a deletion from the N-terminal ends to dsRB-DIII. C, NLS-I, a fragment containing all C-terminal residues after the dsRBDIII. D, dsRBDII, a fragment containing dsRBDII only. E, NoLS-I, a fragment deleting NES and NLS-c. F, NoLS, the fully functional NoLS sequence, containing the P-cluster and dsRBDIII. Bar, 10 m.
actions between the P-cluster and the dsRBD are important for nucleolar binding activity. Nevertheless, the nucleolar localization of ADAR1 isoforms suggests that the nucleolus plays a role in regulating ADAR1-mediated RNA editing.
Because NoLS permits ADAR1 to be recruited to the nucleolus, where actively transcribing rRNA genes and nascent rRNA transcripts are found, the concentration of ADAR1 in this area suggests that ADAR1-mediated RNA editing is involved in the cellular functions of nucleoli. The nucleocapsids of DNA and RNA viruses, which replicate in the nucleus, frequently bear NoLS motifs. Some of these viral proteins appear to participate in the regulation of protein synthesis, perhaps by regulating the synthesis of ribosomal RNA and/or the assembly of ribosomes during viral infection (41). Small amounts of viral nucleocapsid protein in the nucleolus may nonspecifically enhance protein synthesis, whereas high concentrations of these proteins inhibit translation (42). This is consistent with the recent finding that ADAR1 facilitates protein translation in the nucleus or nucleolus (21). The condensation of ADAR1 within the nucleolus may similarly affect nuclear protein synthesis. In addition, the concentration of ADAR1 in the nucleolus may also indicate that it is involved in processing of newly transcribed RNA. Recently, ADAR2 was found to delocalize from the nucleolus and accumulate at sites where the substrate transcripts accumulate (33). Therefore, accumulation of ADAR1 within the nucleolus may also prevent unwanted RNA editing in other cellular compartments, such as pre-mRNA in the nucleus.

Intracellular Distribution of ADAR1 Isoforms Is Regulated through the NES, NoLS, NLS-c, and Regulatory Motifs-We
have demonstrated that multiple ADAR1 isoforms are induced during acute inflammation (30). Interestingly, these isoforms differ in their NES, NoLS, or NLS-c sequences due to alternative splicing (30) or proteolysis. Since NLS-c, NES, and NoLS function independently to control intracellular distribution, the induced ADAR1 isoforms are likely to localize in different cellular compartments. The newly translated ADAR1 is initially imported into the nucleus or retained in the cytoplasm by means of the C-terminal NLS-c. However, the ultimate localization of the imported ADAR1 is dynamically determined by the nucleolar binding activity of NoLS in the middle of the protein and the nuclear exporting activity of NES at the Nterminal end. Eventually, full-length ADAR1 accumulates in the cytoplasm, whereas the short ADAR1 variants lacking the exporter are condensed into the nucleolus and probably associate with nascent rRNA transcripts or actively transcribing rRNA genes within the dense fibrillar components (30). ADAR1 variants lacking the NLS-c will remain in the cytoplasm without undergoing nuclear import and export. Based on these findings, ADAR1 isoforms that have alterations within dsRB-DIII (43) may differ in their nucleolar binding capability because the fully functional NoLS requires dsRBDIII. The isoforms that are truncated from the C-terminal end (44) may not be able to be imported into the nucleus because the NLS-c is deleted. The isoform M296, which is produced through the constitutive promoters (45), may localize within the nucleolus because the NES is missing. Considering that ADAR1 isoforms with differences in their localization signals are differentially induced during acute inflammation (30), it is conceivable that the cellular distribution of ADAR1 isoforms in the cytoplasm, nucleus, or nucleolus is rearranged under certain pathological conditions. This study reveals the complexity of regulatory mechanisms underlying intracellular localization of ADAR1 through nuclear import, export, and nucleolar recruitment. These data suggest that ADAR1-mediated RNA editing occurs in different cellular organelles. Elevated activity of the long form of ADAR1 in vivo could result in highly efficient RNA editing in the cytoplasm (46), supporting the hypothesis that ADAR1-mediated RNA editing occurs in the cytoplasm in addition to the nucleus. Finally, the nucleolar localization of ADAR1 suggests a role for RNA editing in nucleolar function.