RNA Granule Assembly and Disassembly Modulated by Nuclear Factor Associated with Double-stranded RNA 2 and Nuclear Factor 45*

Background: RNA granules are associated with translational control and neurodegenerative disease. Results: Nuclear factor associated with dsRNA 2 (NFAR2) enhances RNA granule assembly through two domains, one of which is antagonized by nuclear factor 45 (NF45). Conclusion: NFAR2 and NF45 play novel regulatory roles in RNA granule assembly and disassembly. Significance: Mechanisms of RNA granule assembly and disassembly are the basis for understanding translational control and neurodegenerative disease. RNA granules are large messenger ribonucleoprotein complexes that regulate translation and mRNA translocation to control the timing and location of protein synthesis. The regulation of RNA granule assembly and disassembly is a structural basis of translational control, and its disorder is implicated in degenerative disease. Here, we used proteomic analysis to identify proteins associated with RNA granule protein 105 (RNG105)/caprin1, an RNA-binding protein in RNA granules. Among the identified proteins, we focused on nuclear factor (NF) 45 and its binding partner, nuclear factor associated with dsRNA 2 (NFAR2), and we demonstrated that NF45 promotes disassembly of RNA granules, whereas NFAR2 enhances the assembly of RNA granules in cultured cells. The GQSY domain of NFAR2 was required to associate with messenger ribonucleoprotein complexes containing RNG105/caprin1, and it was structurally and functionally related to the low complexity sequence domain of the fused in sarcoma protein, which drives the assembly of RNA granules. Another domain of NFAR2, the DZF domain, was dispensable for association with the RNG105 complex, but it was involved in positive and negative regulation of RNA granule assembly by being phosphorylated at double-stranded RNA-activated kinase sites and by association with NF45, respectively. These results suggest a novel molecular mechanism for the modulation of RNA granule assembly and disassembly by NFAR2, NF45, and phosphorylation at double-stranded RNA-activated kinase PKR sites.

RNA granules are large messenger ribonucleoprotein complexes that regulate translation and mRNA translocation to control the timing and location of protein synthesis. The regulation of RNA granule assembly and disassembly is a structural basis of translational control, and its disorder is implicated in degenerative disease. Here, we used proteomic analysis to identify proteins associated with RNA granule protein 105 (RNG105)/ caprin1, an RNA-binding protein in RNA granules. Among the identified proteins, we focused on nuclear factor (NF) 45 and its binding partner, nuclear factor associated with dsRNA 2 (NFAR2), and we demonstrated that NF45 promotes disassembly of RNA granules, whereas NFAR2 enhances the assembly of RNA granules in cultured cells. The GQSY domain of NFAR2 was required to associate with messenger ribonucleoprotein complexes containing RNG105/caprin1, and it was structurally and functionally related to the low complexity sequence domain of the fused in sarcoma protein, which drives the assembly of RNA granules. Another domain of NFAR2, the DZF domain, was dispensable for association with the RNG105 complex, but it was involved in positive and negative regulation of RNA granule assembly by being phosphorylated at double-stranded RNAactivated kinase sites and by association with NF45, respectively. These results suggest a novel molecular mechanism for the modulation of RNA granule assembly and disassembly by NFAR2, NF45, and phosphorylation at double-stranded RNAactivated kinase PKR sites.
RNA granules are large macromolecular complexes composed of mRNA, ribosomes, translation factors, and RNAbinding proteins regulating gene expression by stalling translation (1). RNA granules are also involved in mRNA transport.
Translation is repressed during transport, but it is de-repressed by releasing the mRNA and other components at their destination (2)(3)(4)(5). Thus, RNA granules are in equilibrium between assembly and disassembly to regulate translation, but in neurodegenerative disease such as amyotrophic lateral sclerosis and frontotemporal lobar degeneration, they are prone to conversion into irreversible pathological aggregates (6,7). Recent studies proposed that degenerative disease-associated mutations in RNA granule components, e.g. TAR DNA-binding protein 43 (TDP-43), fused in sarcoma/translocated in sarcoma (FUS 2 /TLS), heterogeneous nuclear ribonucleoprotein (hnRNP) A2B1 and hnRNPA1 enhance their incorporation into RNA granules and promote RNA granule aggregation (6 -8). These proteins contain prionlike low complexity (LC) sequence domains, which are responsible for RNA granule assembly under normal conditions and the formation of pathological aggregates in their mutant forms (6 -9).
Different types of RNA granules have been described, including stress granules (SGs), germ granules, and neuronal RNA granules. SGs are induced by several kinds of stress, such as oxidative stress and virus infections that induce eIF2␣ phosphorylation by heme-regulated eIF2␣ kinase and doublestranded RNA (dsRNA)-activated kinase (PKR), and are implicated in cellular defense against stress (10,11). Neuronal RNA granules are another type of RNA granule that plays central roles in mRNA transport and local translation in dendrites, and they are responsible for synapse formation, plasticity, and long term memory (12)(13)(14). Several RNA-binding proteins are shared between SGs and neuronal RNA granules, e.g. fragile X mental retardation protein, staufen, RasGAP SH3 domain-binding protein (G3BP), and RNA granule protein 105 (RNG105)/caprin1 (1,(15)(16)(17)(18). Expression of RNG105/caprin1 or G3BP that interacts with RNG105/caprin1 (18,19), in cultured A6, 293T, Cos, and HeLa cells, induces the formation of TIA-1-containing SG-like RNA granules in the absence of stressors (18, 20 -22). In neurons, RNG105/caprin1 plays a role in the transport of specific mRNAs into dendrites, and the loss of RNG105/caprin1 results in the degeneration of dendrites and neuronal networks (23). Mice with gene knockouts of RNG105/caprin1 and G3BP exhibit similar phenotypes in terms of fetal growth retardation, cell death in the brain, and neonatal lethality with respiratory failure (23,24).
Nuclear factor associated with dsRNA 1 (NFAR1)/nuclear factor (NF) 90 and NFAR2/NF110 are splice variants transcribed from a single interleukin enhancer binding factor 3 (ILF3) gene (25,26). NFAR1 and NFAR2 were initially identified in a complex with NF45/ILF2 as a transcription activation factor (27,28). Besides transcription regulation, NFAR1 and NFAR2 bind to AU-rich elements in the 3Ј-untranslated region of mRNAs and regulate their stabilization, localization, and translation. For example, NFAR1/2 bind to and stabilize interleukin-2 and mitogen-activated protein kinase phosphatase-1 mRNAs (29,30) and bind to the 3Ј-untranslated region of Tau mRNA, which is known as an element required for its transport to the axon in neurons (31). NFAR1/2 also inhibit global translation as well as bind to viral RNA and prevent its translation (32,33). For the prevention of viral RNA translation, NFAR1/2 are phosphorylated on Thr-188 and Thr-315 by PKR and exported from the nucleus to the cytoplasm to be retained on polysomes (33). NFAR1 and NFAR2 have dsRNA-binding motifs and an Arg-Gly-Gly motif (RGG box) known as an RNA-binding motif. Their zinc finger nucleic acid binding domains (DZF domains) include the PKR phosphorylation sites and form heterodimers with the DZF domain of NF45 (33,34). NFAR2, but not NFAR1, has an additional C-terminal domain, which is enriched with Gly, Gln, Ser, and Tyr residues (GQSY domain), whose function is not well understood except that it can enhance gene expression from specific promoters (35).
In this study, using proteomic analysis, NF45 was identified as a protein associated with RNG105-and G3BP-containing messenger ribonucleoprotein (mRNP) complexes in mitosis when RNA granules disappeared. Expression of NF45 in cultured cells promoted disassembly of RNA granules, whereas its binding partner NFAR2 enhanced RNA granule assembly. Two domains of NFAR2, the GQSY domain similar to the LC sequence domain of FUS and the DZF domain containing PKR phosphorylation sites and that binds to NF45, were involved in the regulation of RNA granule assembly. These results suggest a novel mechanism of RNA granule assembly and disassembly regulated by NFAR2 phosphorylation by PKR and NF45.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-A6 cells were cultured in 10% fetal bovine serum (FBS) and 50% Leivobitz's L-15 medium (Invitrogen) at 23°C without a CO 2 atmosphere. HeLa cells were cultured in 10% FBS and DMEM (Sigma) at 37°C with CO 2 . Mitotic A6 cells were collected selectively by pipetting with a 5-ml plastic pipette after culturing in medium containing 1 g/ml nocodazole for 2 h. Cells remaining on the culture plates after pipetting were used as interphase cells (36). For microscopy experiments, cells were grown on glass coverslips or glass-bottomed dishes. RNG105-GFP and RNG105-monomeric red fluorescent protein 1 (mRFP1) transfectants were produced previously (17,20). Transfections were performed using Lipofectin (Invitrogen) for A6 cells and Lipofectamine 2000 (Invitrogen) for HeLa cells in accordance with the manufacturer's protocols. Transfection of polyinosinic-polycytidylic acid (poly(I-C)) (Sigma) into HeLa cells was performed at a concentration of 2 g/ml. After transfection, cells were cultured for 4 h, then fixed, and immunostained. Stable transfectants were selected in the presence of 0.7 mg/ml geneticin (Invitrogen) followed by picking up fluorescent colonies using a CKX41 microscope (Olympus, Tokyo, Japan) equipped with an epifluorescence module. Sodium arsenite (Wako Pure Chemical, Osaka, Japan) was added to culture medium at concentrations indicated in the figure legends. Imidazolo-oxindole PKR inhibitor C16 (Sigma) was added at 1 M.
Immunoprecipitation (IP)-IP was performed using agaroseconjugated anti-GFP (Medical and Biological Laboratories, Nagoya, Japan) as described previously (17). Briefly, cells were homogenized in 0.25 M sucrose, 0.7% Triton X-100, 0.1 mM DTT, protease inhibitors (10 g/ml leupeptin, pepstatin, aprotinin, and 1 mM PMSF), and 1,000 units/ml RNase inhibitor (Takara Bio, Shiga, Japan), and then centrifuged at 10,000 ϫ g for 10 min at 4°C. The supernatant was added to 1:10 volume of 10ϫ PBS followed by 1:20 volume of anti-GFP-agarose beads. After rocking for 2 h at 4°C, the beads were washed three times in PBS containing 0.1 mM DTT, protease inhibitors, and 100 units/ml RNase inhibitor. IP in the presence of RNase was performed in the continuous presence of 0.2 mg/ml RNase A (Wako Pure Chemical) without RNase inhibitor in the cell extracts and the wash buffer. Immunoprecipitates were analyzed by SDS-PAGE using a two-dimensional Silver Stain II kit (Cosmo Bio, Tokyo, Japan), Western blotting, or mass spectrometry.
Mass Spectrometry-Immunoprecipitates with the anti-GFP antibody were separated by SDS-PAGE and stained using the Silver Stain MS kit (Wako Pure Chemical). After bands were cut out from the gel, they were destained with 15 mM K 3 (Fe(CN) 6 ) and 50 mM Na 2 S 2 O 3 for 10 min, washed with H 2 O, dehydrated with 50% acetonitrile in 25 mM NH 4 HCO 3 for 5 min, and dried in a vacuum desiccator. The gel slices were deoxidized in 10 mM DTT in 25 mM NH 4 HCO 3 at 56°C for 1 h, washed with 25 mM NH 4 HCO 3 , alkylated with 55 mM iodoacetamide in 25 mM NH 4 HCO 3 at room temperature for 45 min, dehydrated, and dried again. After the gel slices were rehydrated with 10 g/ml trypsin in 50 mM NH 4 HCO 3 on ice for 30 min, excess solution was removed, and the gel slices were incubated at 37°C for 12 h for in-gel digestion. Digested peptides were extracted with 50% acetonitrile and 5% CF 3 COOH at room temperature for 1 h. After freeze-drying, the peptides were dissolved in 30% acetonitrile and 0.1% formic acid and then analyzed with Q-TOF Premier (Waters, Milford, MA).
Fluorescence Microscopy and Image Analysis-Fluorescence images were acquired using a DeltaVision optical sectioning microscope (Applied Precision, Issaquah, WA) equipped with an IX70 microscope (Olympus) with a PlanApo 60ϫ or a UApo 40ϫ oil objective lens or using an A1 confocal microscope equipped with a Ti-E inverted microscope (Nikon Instruments, Melville, NY) with a PlanApo VC60ϫ oil objective lens. Timelapse imaging of A6 cells was performed using the optical sectioning microscope with the 40ϫ oil objective lens at room temperature.
The images were converted into binary images and quantified for the area of cells, nuclei, and RNA granules using Adobe Photoshop software. To analyze the fluorescence intensity of RNA granules and nuclei, original images were multiplied by the binary images and then the mean pixel intensity in each area was measured using Photoshop software. To analyze the fluorescence intensity of the cytoplasm, original images were multiplied by inverted binary images to mask nuclei and RNA granules, and then the mean pixel intensity was measured. Mean background fluorescence was subtracted from the mean intensity of the target areas.
Sucrose Density Gradient Centrifugation-Cultured cells were incubated with 100 g/ml cycloheximide in culture medium for 15 min, washed with ice-cold PBS, and then lysed with cell lysis buffer (20 mM Hepes-KOH, pH 7.4, 15 mM MgCl 2 , 200 mM KCl, 1% Triton X-100, 100 g/ml cycloheximide, 2 mM DTT, 1 mg/ml heparin, and protease inhibitors). After centrifugation at 14,000 ϫ g for 5 min at 4°C, the supernatant was overlaid onto 7-47% w/w linear sucrose gradients in the cell lysis buffer lacking Triton X-100 and heparin, which were prepared using a gradient gel-making device (ATTO, Tokyo, Japan) and then centrifuged at 100,000 ϫ g for 4 h at 4°C in an SW41Ti swing rotor (Beckman Coulter, Brea, CA). The sample was fractionated in 20 fractions, and their absorbance at 254 nm was measured using NanoDrop (Thermo Fisher Scientific, Waltham, MA).
Cell Survival Analysis-Cultured cells were treated with several concentrations of arsenite for 3 h. After washing twice with the culture medium, the cells were cultured for an additional 21 h in the absence of arsenite. Phase contrast images of the cells were acquired using the CKX41 microscope with an LCAch N 20ϫ objective lens and an E-620 digital single lens reflex camera (Olympus) to count the number of cells attached to and detached from the culture plates. Detached cells were collected by pipetting and treated with 0.2% trypan blue in PBS for 15 min to stain nonviable cells. Viable and nonviable cells were counted using the CKX41 microscope.
Ribopuromycilation Assay-Ribopuromycilation assay was conducted as described previously with a modification (22). Briefly, cells were pulse-labeled with 50 g/ml puromycin and 100 g/ml cycloheximide for 10 min at 23°C and washed with 100 g/ml cycloheximide in PBS for 3 min on ice. After permeabilization with 50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 25 mM KCl, 100 g/ml cycloheximide, 10 units/ml RNase inhibitor, protease inhibitors, and 0.015% digitonin for 1 min on ice, the cells were fixed by adding 1:10 volume of ice-cold 37% form-aldehyde for 5 min on ice. After postfixation with 3.7% formaldehyde in PBS for 10 min at room temperature, the cells were subjected to immunofluorescence staining with the anti-puromycin antibody.
Statistical Analysis and Informatics-Data are expressed as mean Ϯ S.E. Two samples were compared using an unpaired t test. More than two samples were compared using one-way analysis of variance followed by Tukey-Kramer test. Pearson's product-moment correlation coefficients were calculated to measure the relationship between two variables. 2 test was used to test the difference in the proportion of living and dead cells between two samples.
Prediction of protein structural disorder was performed using the DISOPRED2 predictor (37) and the PONDR-FIT and VSL2B predictors (38,39). Disorder-based potential binding sites were predicted using the ANCHOR and the MoRFpred predictors (40,41).

Identification of NF45 in mRNP Complexes
Containing RNG105 and G3BP-Using anti-GFP antibody-conjugated beads, we performed IP from A6 cells stably expressing RNG105-GFP or G3BP-GFP. SDS-PAGE of the immunoprecipitates showed that the GFP fusion proteins and their associated proteins were immunoprecipitated from the transfectants (Fig. 1A). Analysis of these proteins by mass spectrometry identified several proteins already known to be localized to RNA granules and other proteins, including RNA-binding proteins, proteasome-associated proteins, chaperon proteins, and proteins involved in methylation (Fig. 1D). The identified proteins were fused to GFP and expressed in A6 cells with RNG105-mRFP1, which found proteins that co-localized with RNA granules and protein N-methyltransferase 1 that induced the disassembly of RNG105-localized RNA granules (Fig. 1, D and E).
During the observation of A6 cells expressing RNG105-GFP or G3BP-GFP by fluorescence microscopy, we noticed that RNA granules disappeared in mitotic cells. Time-lapse observation of A6 cells co-expressing RNG105-mRFP1 and G3BP-GFP revealed that RNG105 and G3BP were co-localized in RNA granules in interphase cells, but these RNA granules disappeared within a few minutes after nuclear envelope breakdown and reappeared after nuclear formation following cell division (Fig. 1B). This led to the idea that the protein composition of mRNP complexes containing RNG105 and G3BP may change between interphase and mitotic cells. IP of RNG105-GFP or G3BP-GFP from RNG105-GFP-or G3BP-GFP-expressing interphase and mitotic cells showed that the SDS-PAGE pattern of the immunoprecipitates was largely unchanged, but a few proteins changed in abundance between interphase and mitotic cells (Fig. 1C). A major change was an increase in the amount of ϳ45-kDa proteins in the mitotic immunoprecipitates both from RNG105-GFP-and G3BP-GFP-expressing cells (Fig. 1C). Mass spectrometric analysis of these proteins identified them as NF45 (Fig. 1D). Western blotting of the immunoprecipitates confirmed that the association of NF45 with the RNG105-GFP mRNP complex was increased in mitotic cells, which was because of the increase of NF45 in input extracts (Fig. 1F). Because the total amount of NF45 was not increased in mitosis, the increase may be caused by release of NF45 from the nucleus after nuclear envelope breakdown in mitosis.
NF45 Disassembles RNA Granules and Renders Cells Susceptible to Stress-Expression of NF45 as a GFP fusion protein in A6 cells with RNG105-mRFP1 showed that NF45 induced the disassembly of RNG105-localizing RNA granules (Fig. 1,  D and E). Similarly, expression of NF45-mRFP1 in A6 cells stably expressing RNG105-GFP resulted in the disassembly of RNG105-localizing RNA granules (Fig. 2, A and E). The extent of disassembly was dependent on the expression level of NF45 (Fig. 2D). Expression of NF45-mRFP1 also induced the disassembly of G3BP-GFP RNA granules, but expression of control mRFP1 did not (Fig. 2, B, C, and F). These results indicated that high level expression of NF45 disassembled RNA granules containing RNG105 and G3BP.
Because RNA granules (SGs) are implicated in cellular defense against stress, we examined whether expression of NF45 affected cell resistance to stress. First, A6 cells stably expressing NF45-GFP were generated, and the effect of NF45 expression on stress-induced SG formation was analyzed. In control A6 cells, SGs were induced by arsenite treatment, as judged by immunostaining with an anti-RNG105 antibody. In contrast, SG formation was inhibited in NF45-GFP-expressing cells (Fig. 3, A and B). Inhibition of SG formation by transient NF45 expression was also observed in time-lapse imaging of RNG105-GFP-expressing cells after arsenite treatment (Fig. 4). The inhibition of SG formation by NF45 was not mediated by restoring arsenite-induced translation inhibition because polysome profiles after arsenite treatment were indistinguishable between control and NF45-expressing cells (Fig. 3C). Next, NF45-GFP stable transfectants and control cells were cultured in the presence of arsenite for 3 h and then further cultured in the absence of arsenite (Fig. 3D). Most of the cells were rounded up after treatment with arsenite at concentrations higher than 1 mM, but some cells survived and reattached to the culture plate after culturing without arsenite. Expression of NF45 reduced the number of cells that reattached to the culture plate (Fig. 3, D and E). Furthermore, the number of dead cells was increased by NF45 expression (Fig. 3E). These results indicated that the expression of NF45 rendered cells susceptible to arsenite stress and suggested the link between SG assembly and cellular defense against stress.
NFAR2, but Not NFAR1, Localizes with and Enlarges RNA Granules-Because NF45 forms heterodimers with NFAR1 and NFAR2 (27,42), we investigated whether NFAR1 and NFAR2 also have effects on RNA granule assembly. NFAR1 and NFAR2 were fused to GFP and expressed with RNG105-mRFP1 in A6 cells. NFAR1 was located in the nucleus and was not concentrated in the RNG105-localizing RNA granule area (Fig.  5, B and C). In contrast, NFAR2 was located predominantly in the nucleus but also co-localized with RNG105 RNA granules (Fig. 5, B and C). These results indicated that the C-terminal GQSY domain of NFAR2 was required for its localization with RNA granules because only NFAR2, but not NFAR1, has the GQSY domain (Fig. 5A). In addition to co-localization with RNA granules, NFAR2 increased the size of RNG105 RNA granules (Fig. 5, B and C). IP analysis of transfectants using an anti-GFP antibody showed that RNG105-mRFP1 was co-immunoprecipitated with NFAR2-GFP but not with NFAR1-GFP (Fig. 5, D and E). RNase treatment reduced the amount of RNG105-mRFP1 co-precipitates but still left small interaction with NFAR2-GFP (Fig. 5, D and E), suggesting that NFAR2 can associate with RNG105-containing mRNP complexes through protein-protein interaction, and the mRNP complexes may be disassembled by RNase treatment. Taken together, these results suggested that NFAR2 is associated with RNG105-localizing RNA granules through the GQSY domain and enhances the assembly of RNA granules.
GQSY Domain of NFAR2 Is Similar to the LC Sequence Domain of FUS-We then analyzed the structural and functional features of the GQSY domain. The amino acid composition of the GQSY domain has little diversity, i.e. the domain is highly enriched with Gly, Gln, Ser, and Tyr residues, but other Arrows indicate proteins whose amounts were higher in mitotic co-immunoprecipitates than in interphase. D, gel regions indicated in A and C were cut and analyzed by mass spectrometry. Identified proteins were classified according to their functions. Numbers correspond to the region numbers in A and C. The proteins were fused to GFP and expressed in A6 cells with RNG105-mRFP1 to test their co-localization with RNA granules and effects on RNA granules. The results are indicated on the right side. G, co-localized with RNA granules; D, promotes disassembly of RNA granules; NE, no effect on RNA granules; NT, not tested. E, representative images of A6 cells co-expressing the identified proteins fused to GFP and RNG105-mRFP1. Scale bar, 10 m. F, input extracts, and the immunoprecipitates with the anti-GFP antibody from total, interphase (I), or mitotic (M) A6 cells expressing RNG105-GFP were immunoblotted with anti-GFP and anti-NF45 antibodies. Extracts with SDS-PAGE sample buffer were also blotted. Bottom panels show quantification of the band intensity of immunoblotting (n ϭ 3). charged or nonpolar amino acids are infrequent (Fig. 6A). Because such LC sequences are suggested to characterize natively disordered polypeptides (43), the NFAR2 sequence was subjected to predictive analysis of native disorder. The disorder probability of the GQSY domain was high (Fig. 6B), supporting the notion that the GQSY domain includes an LC sequence. In addition, the GQSY domain was predicted to contain multiple regions that bind to protein partners through disorder-based binding (Fig. 6B). It has been reported that the LC sequence domain of FUS can undergo phase transition from a soluble phase to a hydrogel-like state in vitro, and the domain is required for localization to SGs in cells (9,44). The FUS LC sequence domain contains 27 variants of tripeptide sequence (G/S)Y(G/S) repeats, which are required for hydrogel formation and localization to SGs (9,44). Comparison of the amino acid compositions of the NFAR2 GQSY domain and the FUS LC sequence domain revealed that the two domains share a high level of similarity (Fig. 6A). Furthermore, the NFAR2 GQSY domain contains 24 variants of (G/S)Y(G/S) repeats (Fig.  6C). Thus, the GQSY domain of NFAR2 is structurally similar to the LC sequence domain of FUS.
Next, we examined whether the GQSY domain is functionally similar to the LC sequence domain of FUS. Full-length FUS, LC sequence domain deletion mutant (FUS⌬LC), and a chimeric FUS with its LC sequence domain replaced by the NFAR2 GQSY domain (GQSY-FUS⌬LC) were constructed (Fig. 6D). These constructs were fused to GFP and expressed in A6 cells with RNG105-mRFP1 (Fig. 6E). FUS co-localized with RNG105 and increased the size of RNG105-localizing RNA granules (Fig. 6, E and F). In contrast, although FUS⌬LC co-localized with RNG105, it lost the ability to increase the size of RNA granules, indicating that the LC sequence domain was required for the assembly of large RNA granules (Fig. 6, E and F). Expression of GQSY-FUS⌬LC resulted in large RNG105 RNA granules, which indicated that the GQSY domain restored the ability to assemble large RNA granules (Fig. 6, E and F). The restoration was not because of higher expression of GQSY-FUS⌬LC than FUS⌬LC or increased expression of RNG105 in the GQSY-FUS⌬LC transfectants (Fig. 6G). These results indicated that the GQSY domain of NFAR2 was functionally exchangeable with the LC sequence domain of FUS in terms of the ability to assemble large RNA granules.
DZF Domain of NFAR2 Is Required to Enhance RNA Granule Assembly-The DZF domains of NFAR1 and NFAR2 are phosphorylated at Thr-188 and Thr-315 by PKR, which causes their retention on ribosomes and translational inhibition (Fig. 7A) (33). In addition to being phosphorylated by PKR, the DZF domain is bound by NF45 to form a heterodimer (Fig. 7A) (27,42). We then examined the effect of the DZF domain of NFAR2 on RNA granule assembly. We constructed NFAR2 lacking the DZF domain (NFAR2⌬DZF), fused it to GFP, and introduced it to A6 cells with RNG105-mRFP1 (Fig. 7, A and B). The distribution pattern of NFAR2⌬DZF was similar to that of wild-type NFAR2, and it co-localized with RNG105 RNA granules (Fig. 7,  B and C). IP from the transfectants also indicated that NFAR2⌬DZF-GFP associated with the RNG105-mRFP1 complex at the same level as wild-type NFAR2-GFP (Fig. 7E). However, the expression of NFAR2⌬DZF did not induce large RNG105-containing RNA granules, as compared with wild-type NFAR2, which suggested that the DZF domain was required to enhance the assembly of RNA granules (Fig. 7, B and D). Together, these results suggested that the DZF domain is involved in the assembly of RNA granules without affecting the association between the GQSY domain and RNG105 mRNP complexes.   . Arrows indicate RNG105-mRFP1 granules. The brightness of pictures was adjusted to show RNA granules distinctly, and the original images around nuclei used to quantify fluorescence intensity in C are shown in the insets. Scale bar, 10 m. C, quantification of RNA granules in B. Fluorescence intensity of NFAR1-GFP and NFAR2-GFP in granules was normalized to that in the nucleus of the same cell. NFAR2-GFP was co-localized with granules and increased the mean size and the total area of granules in the cytoplasm. n ϭ 30 cells; ***, p Ͻ 0.001, t test. D, IP from A6 cells co-expressing RNG105-mRFP1 and NFAR1-GFP, NFAR2-GFP, or GFP with the anti-GFP antibody in the presence or absence of RNase. Immunoprecipitates were immunoblotted with anti-GFP and anti-RFP antibodies. E, quantification of the band intensity in D. n ϭ 3.

Phosphomimetic Mutations of NFAR2 at PKR Sites in the DZF Domain Enhances Cytoplasmic Retention, RNA Granule
Localization, and Assembly-To examine the effect of NFAR2 phosphorylation by PKR on its localization to, and the assembly of, RNA granules, phosphodeficient (TA) and phosphomimetic (TD) mutants of NFAR2 were constructed and expressed as GFP fusion proteins in A6 cells (Figs. 7E, 8, and 9A). When the NFAR2-TA mutant was expressed with RNG105-mRFP1, NFAR2-TA reduced its co-localization with RNG105 and also reduced the size of RNG105 RNA granules compared with wild-type NFAR2 (Fig. 8, A-D). In contrast, the NFAR2-TD mutant increased its co-localization with RNG105 and induced larger RNA granules than wild-type NFAR2 (Fig. 8, A-D). Because the number of RNG105-localizing RNA granules was not increased by expression of wild-type NFAR2 or NFAR2-TD, the increase in the total area of RNA granules was attrib-uted to the increase in the size of RNA granules (Fig. 8, C and  D). Quantitative analysis showed that increased localization of NFAR2-TD to RNA granules was correlated with the increase in its cytoplasmic localization, and reduced localization of NFAR2-TA to RNA granules was correlated with its reduced localization in the cytoplasm (Fig. 8E). These results suggested that phosphorylation of the DZF domain of NFAR2 by PKR increases its cytoplasmic retention, which increases its association with RNA granules to enhance the assembly of RNA granules.
Next, we tested whether the increase in the cytoplasmic and RNA granule localization of NFAR2 was dependent on PKR activity. It is reported that PKR is activated by diverse stress signals, including arsenite through PKR activating protein (PACT) (45, 46), although heme-regulated eIF2␣ kinase is a more favorable target of arsenite (47). PKR is more strongly activated by dsRNA (poly(I-C)), which is demonstrated to phosphorylate NFAR1 and NFAR2 (33). In addition, PKR is activated by G3BP-induced RNA granule formation (22). First, we examined the effect of poly(I-C)-induced PKR on NFAR2 localization to RNA granules (SGs). In this analysis, HeLa cells were used because poly(I-C) transfectants were not distinguished from nontransfectants, and transfection efficiency of HeLa cells was much higher (Ͼ90%) than A6 cells (Ͻ20%). Without poly(I-C), SGs were not apparently formed in the cytoplasm as detected by immunostaining for G3BP (Fig. 10A). When poly(I-C) was introduced in cells, it induced the formation of G3BP-containing SGs, which were immunostained with an anti-NFAR1/2 antibody. SG formation by poly(I-C) was reduced by PKR inhibition, and concomitantly, SG staining with the NFAR1/2 antibody was reduced (Fig. 10, A and B). Thus, NFAR1/2 was localized to PKR-induced SGs.
Next, we reverted to using the A6 transfectants and examined the effect of PKR inhibition on NFAR2 localization to the cytoplasm and RNA granules. As reported previously (18), expression of RNG105 increased the phosphorylation level of eIF2␣ (Fig. 11, A and B). This phosphorylation was reduced by PKR inhibition, indicating that PKR was activated by RNG105 expression (Fig. 11, A and B), which was similar to G3BP expression (22). NFAR2 co-expressed with RNG105 was localized mainly in the nucleus and also in the cytoplasm and RNA granules (Fig. 11C). However, when PKR was inhibited, NFAR2 localization to the cytoplasm and RNA granules was markedly reduced, and it was restricted to the nucleus (Fig. 11, C and D). In contrast, localization of NFAR2-TD in the cytoplasm and RNA granules was not changed even when PKR was inhibited (Fig. 11, C and D). These results indicated that the increase of NFAR2 in the cytoplasm and RNA granules was PKR-dependent and that Thr-188 and Thr-325 were the target sites of PKR. Although the localization of NFAR2-TD was not affected by PKR inhibition, RNA granule formation was impaired by PKR inhibition (Fig. 11, C and E). This may be because other endogenous pathways necessary for the formation RNA granules, e.g. eIF2␣ phosphorylation, were inhibited. However, NFAR2-TD still induced larger RNA granules than wild-type NFAR2 or NFAR2-TA even in the presence of PKR inhibitor, suggesting that phosphomimetic mutations overcame the effect of PKR inhibition (Fig. 11, C and E). Taken together, these results supported the notion that NFAR2 phosphorylation by PKR enhances its cytoplasmic retention, localization to RNA granules, and the assembly of RNA granules.
NF45 Antagonizes RNA Granule Assembly by NFAR2-It is known that NF45 binds directly to NFAR1/2 (27,42). Co-expression of NFAR2-GFP and NF45-mRFP1 in A6 cells and subsequent IP with an anti-GFP antibody confirmed that NFAR2 and NF45 formed a complex independently on RNA (Fig. 9A). The amount of NF45 bound to NFAR2 was not significantly changed by mutations in the PKR phosphorylation sites (Fig.   FIGURE 7. NFAR2⌬DZF co-localizes with RNG105 RNA granules but loses the ability to enlarge RNA granules. A, schematic diagram of NFAR2⌬DZF. The DZF domain includes PKR phosphorylation sites (Thr-188 and Thr-315) and binds to NF45. B, A6 cells were co-transfected with RNG105-mRFP1 and NFAR1-GFP, NFAR2-GFP, or NFAR2⌬DZF-GFP. Arrows indicate RNG105-mRFP1 granules. Scale bar, 10 m. C, quantification of cytoplasmic and granule localization of NFAR1-GFP, NFAR2-GFP, and NFAR2⌬DZF-GFP in B. D, quantification of the mean size and total area of RNG105-mRFP1 granules in B. n ϭ 25; *, p Ͻ 0.05; **, p Ͻ 0.01, Tukey-Kramer test. E, IP from A6 cells co-expressing RNG105-mRFP1 and NFAR2-GFP, NFAR2-TA-GFP, NFAR2-TD-GFP, NFAR2⌬DZF-GFP, or GFP using the anti-GFP antibody in the presence or absence of RNase. Immunoprecipitates were immunoblotted with anti-GFP and anti-RFP antibodies. Bottom panels show quantification of the band intensity (n ϭ 3). 9A). Next, we examined the effect of NF45 binding to NFAR2 on RNA granule assembly. A6 cells were co-transfected with RNG105-mRFP1 and NFAR2-GFP or its phosphorylation site mutants and additionally with or without NF45-mRFP1 (Fig. 8,  A and F). Expression of NF45 changed the distribution pattern of wild-type NFAR2 and NFAR2-TA, i.e. their cytoplasmic localization was increased by NF45 expression (Fig. 8, A, F, and  G). This may be because cytoplasmic localization of NF45-mRFP1 was higher than that of NFAR2-GFP or NFAR2-TA-GFP (cytoplasmic NF45-mRFP1 was 47.7 Ϯ 3.6% of nuclear NF45-mRFP1; cf. Fig. 8G) and was accompanied by NFAR2-GFP or NFAR2-TA-GFP that bound to NF45-mRFP1. Concomitantly, their localization to RNG105 RNA granules was also increased in NF45-expressing cells (Fig. 8G). In contrast, the distribution pattern of NFAR2-TD was not affected by NF45 expression (Fig. 8, A, F, and G). This may be because cytoplasmic localization of NFAR2-TD was relatively high even without NF45 expression, or NF45 may have little effect on the localization of phosphorylated NFAR2.
Although cytoplasmic and RNA granule localization of wildtype NFAR2 and NFAR2-TA was increased by NF45 co-expres-sion, the assembly of RNG105-containing RNA granules was reduced by NF45 co-expression (Fig. 8, A, F, and H). NF45 expression also reduced the enhancement of RNA granule assembly by NFAR2-TD (Fig. 8, A, F, and H). These results suggested that NF45 antagonized the ability of NFAR2 to enhance RNA granule assembly in the cytoplasm. Nevertheless, RNA granule assembly was enhanced more by NFAR2-TD than by wild-type NFAR2 or NFAR2-TA, even when NF45 was coexpressed (Fig. 8, F and H). Thus, RNA granule assembly was balanced by negative regulation by NF45 binding to NFAR2 and positive regulation by phosphorylation at PKR sites on NFAR2. Phosphorylation of NFAR2 contributes not only to its cytoplasmic retention but also to increasing its activity to assemble large RNA granules by overcoming NF45 in the cytoplasm.
Although NF45 binding to NFAR2 reduced the assembly of RNG105-containing RNA granules, IP analysis indicated that NF45 binding to NFAR2 did not change the association between NFAR2 and the RNG105 mRNP complex (Fig. 9B). Similarly, although phosphorylation site mutations in NFAR2 influenced the assembly of RNG105-containing RNA granules, these mutations did not change the association between A, both Thr-188 and Thr-315 in the DZF domain of NFAR2 were mutated to Ala or Asp in phosphodeficient (TA) or phosphomimetic (TD) mutants, respectively. A6 cells were co-transfected with RNG105-mRFP1 and NFAR2-GFP, NFAR2-TA-GFP, or NFAR2-TD-GFP. Scale bar, 10 m. B-D, quantification of RNG105-mRFP1 granules in A. The TA mutant reduced its ability to localize to and enhance the assembly of RNG105 granules, whereas the TD mutant increased the ability compared with wild-type (WT) NFAR2. n ϭ 25; **, p Ͻ 0.01, Tukey-Kramer test. E, correlation between cytoplasmic localization and granule localization of NFAR2-GFP and its mutants. r ϭ 0.87, p ϭ 2.68 ϫ 10 Ϫ20 . F, same experiments as A were performed except that the cells were additionally co-transfected with NF45-mRFP1. Although RNG105 and NF45 were visualized in the same color, RNG105 granules were distinguishable because of their bright signals. Essentially the same results were obtained by using NF45 without fluorescent protein tags. Scale bar, 10 m. G, quantification of cytoplasmic and granule localization of NFAR2-GFP and its mutants in A and F. H, quantification of the mean size and total area of RNG105-mRFP1 granules in A and F. n ϭ 25; **, p Ͻ 0.01; ***, p Ͻ 0.001, t test between NF45-mRFP1-transfected and -untransfected cells. #, p Ͻ 0.05; ##, p Ͻ 0.01, Tukey-Kramer test among NF45-mRFP1-transfected cells.
NFAR2 and the RNG105 mRNP complex (Fig. 7E). These results were consistent with the results that the deletion of the DZF domain did not change the association between NFAR2 and the RNG105 mRNP complex (Fig. 7E). Taken together, these results suggested that the DZF domain regulates RNA granule assembly positively and negatively by being phosphorylated at PKR sites and by association with NF45, respectively, not through regulating the interaction between the GQSY domain and the RNG105 mRNP complex.
Translational Repression by NFAR1/2 and De-repression by NF45-RNA granule assembly and disassembly are often linked to translational repression and activation. Then we examined the effect of the expression of NFAR1, NFAR2, and NF45 on translation in cells. Because stable transfectants of NFAR1 or NFAR2 were not obtained, we conducted ribopuromycilation assay (22) instead of sucrose density gradient centrifugation. Treatment of cells with arsenite reduced anti-puromycin antibody staining (Fig. 12, A and B), which confirmed the reproducibility of this method. Expression of NFAR1, NFAR2, or its mutants significantly reduced the staining with the antibody, suggesting that NFAR2 inhibited translation independently on its ability to localize to RNA granules (Fig. 12, C and D). Co-expression of NF45 antagonized the translational repression by NFAR1 and NFAR2 (Fig. 12, C and D), suggesting linkage between the roles of NF45 in RNA granule disassembly and translational de-repression. But the translational de-repression by NF45 may be rather specific for NFAR1 and NFAR2 because polysome dissociation induced by arsenite, which is mediated FIGURE 9. Binding of NF45 to NFAR2 does not affect the association between NFAR2 and RNG105 mRNP complexes. A, IP from A6 cells co-expressing NF45-mRFP1 and NFAR2-GFP, NFAR2-TD-GFP, NFAR2-TA-GFP, or GFP using the anti-GFP antibody in the presence or absence of RNase. Immunoprecipitates were immunoblotted with anti-GFP and anti-RFP antibodies. Lower bands of NF45-mRFP1 were detected in cell lysates (input) and the NFAR2-GFP immunoprecipitates, although the reasons are not known. B, IP with the anti-GFP antibody from A6 cells co-expressing RNG105-mRFP1 and NFAR2-GFP or GFP, which were additionally transfected with or without NF45-mRFP1. C, a model for the role of the GQSY and DZF domains of NFAR2 and NF45 in RNA granule assembly. by eIF2␣ phosphorylation, was not affected by NF45 (Fig. 3C). Thus, RNA granule assembly and disassembly modulated by NFAR2 and NF45 were not likely to be strictly linked to translation regulation, providing a novel regulatory mechanism for RNA granule assembly and disassembly. Nevertheless, as NFAR2 and eIF2␣ are both the targets of PKR, RNA granule assembly by NFAR2/NF45 and translation initiation are simultaneously regulated under the control of PKR.

DISCUSSION
This study identified NFAR2 and NF45 as positive and negative regulatory proteins, respectively, of RNA granule assembly. Although it has been unclear whether the two NFAR splice variants, NFAR1 and NFAR2, have different properties from one another, we demonstrated that NFAR2, but not NFAR1, has the ability to localize to and enhance the assembly of RNA granules. This ability is attributed to the C-terminal GQSY domain, which interacts with mRNP complexes containing RNG105 and has similar structural and functional features to the FUS LC sequence domain. Another domain of NFAR2 responsible for the assembly of RNA granules is the DZF domain, whose ability to enhance RNA granule assembly is balanced by positive regulation by phosphorylation at PKR sites and negative regulation by binding with NF45. The results provide new insights into the molecular mechanism by which NFAR2, its phosphorylation by PKR, and NF45 regulate RNA granule assembly.
Previous proteomic analyses suggested that NFAR1 and/or NFAR2 were contained in mRNP complexes. IP of RNA granule components such as Argonaute proteins or TDP-43 identified NFAR1/2 together with other RNA granule components such as fragile X mental retardation protein, HuR, poly-A-binding protein, staufen, FUS, G3BP, and RNG105/caprin1 (48,49). However, it remained unclear whether NFAR proteins are in fact components of RNA granules. We demonstrated that NFAR2, but not NFAR1, was a component of RNA granules.
NFAR2, but not NFAR1, has the GQSY domain that is similar to the FUS LC sequence domain, a subtype of intrinsically disordered regions. Intrinsically disordered proteins often bind to their specific partners via disorder-to-order transition, playing roles in regulating the function of their binding partners and in promoting the assembly of macromolecular complexes (50). Intrinsically disordered regions are commonly found in proteins that nucleate SGs and other SG-associated proteins involved in cellular signaling (51). Intrinsically disordered proteins are also associated with numerous human diseases, including neurodegeneration and cancer (50).
The phase transition of FUS LC sequence domains from a soluble phase to a hydrogel-like state is proposed to be the molecular principle driving RNA granule assembly (9,44), which is consistent with our results that FUS and GQSY-FUS⌬LC, but not FUS⌬LC, increased the size of RNA granules (Fig. 6). Serine residues in the (G/S)Y(G/S) motif of FUS LC sequence domain are reportedly phosphorylated by DNA-dependent protein kinase (DNA-PK), which reduces the binding of FUS to the hydrogel, offering a means of dynamic control of RNA granule assembly (44). NFAR1/2 are also reported to form a complex with and are phosphorylated by DNA-PK in the nucleus (52). Because RNA granules are formed not in the nucleus but in the cytoplasm, it will be of interest to investigate in the future whether the phosphorylation of NFAR2 by DNA-PK controls RNA granule assembly and disassembly in the cytoplasm and nucleus.
The GQSY domain was required for the association with RNG105 mRNP complexes, and the binding of RNG105 to NFAR2 included both RNase-resistant and sensitive fractions. This suggests that a fraction of RNG105 was associated with the GQSY domain of NFAR2 through protein-protein interactions, and other fractions were associated in an RNA-dependent manner. RNG105 mRNP complexes may contain multiple RNG105 molecules and other proteins, which are bridged by mRNAs, and associate with the GQSY domain of NFAR2 (Fig. 9C).
Immunoprecipitates with RNG105 from mitotic cells included multiple types of proteins whose composition was largely unchanged from interphase cells, although RNA granules disappeared in mitotic cells (Fig. 1, B and C). These results suggest that small RNG105 mRNP complexes are maintained even when large RNA granules are disassembled in mitosis. We hypothesize that RNG105 mRNP complexes may exist as small units, and the units are connected to each other when RNA granules are formed. The connection may be released by proteins like NF45 because the association of NF45 with RNG105 mRNP complexes was increased in mitosis, and expression of NF45 in cells led to disassembly of RNA granules. Because the FIGURE 11. PKR inhibition reduces the cytoplasmic and RNA granule localization of NFAR2, but not of NFAR2-TD. A, A6 cells co-transfected with RNG105-mRFP1 and NFAR2-GFP were cultured in the presence or absence of PKR inhibitor and stained with an anti-phospho-eIF2␣ antibody (Ab). Arrows and arrowheads indicate transfected and untransfected cells, respectively. Scale bar, 10 m. B, quantification of phospho-eIF2␣ fluorescence intensity in transfectants expressing RNG105-mRFP1 with or without NFAR2-GFP or its mutants. Fluorescence intensity was normalized to that in untransfected neighboring cells. C, A6 cells co-expressing RNG105-mRFP1 and NFAR2-GFP or its mutants were cultured without (left panels) or with (right panels) PKR inhibitor. Arrows indicate cytoplasmic and granule localization of NFAR2-GFP, which was reduced by PKR inhibition. Arrowheads indicate cytoplasmic and granule localization of NFAR2-TD-GFP, which was not significantly affected by PKR inhibition. Scale bar, 10 m. D, quantification of cytoplasmic and granule localization of NFAR2-GFP and its mutants in C. E, quantification of granules in the cells in C. n ϭ 30; ***, p Ͻ 0.001, t test between control and PKR inhibitor-treated cells. ##, p Ͻ 0.01, Tukey-Kramer test among PKR inhibitor-treated cells.
increased association of NF45 with RNG105 mRNP complexes was likely because of the release of NF45 from the nucleus in mitosis, its binding partner NFAR2 may also be released from the nucleus and increase the association with RNG105 mRNP complexes in mitosis. The NF45-NFAR2 dimer could compete with other factors for the binding to RNG105 mRNP complexes and induce RNA granule disassembly, which may be supported by the results that co-expression of NF45 and NFAR2 led to disassembly of RNG105-induced RNA granules (Fig. 8F, cf. 5B, control).
The DZF domain of NFAR2 was dispensable for the association with RNG105 mRNP complexes but responsible for RNA granule assembly. This suggests that the DZF domain forms complexes independently on the RNG105 mRNP complex. Formation of the DZF domain-including complex may be enhanced by phosphorylation with PKR and inhibited by binding with NF45, neither of which change the association between NFAR2 and RNG105 (Figs. 7E, and 9, B and C). NFAR2 may function as a connector of RNG105 mRNP complexes through its multivalent domains as follows: the GQSY domain that associates with RNG105 mRNP complexes and the DZF domain. When the DZF domain is bound by NF45, the connector function could be reduced, which may explain the mechanism of how NF45 disassembles RNA granules. This model does not exclude another possibility that proteins such as FUS also function as connector proteins, proteins such as protein N-methyltransferase 1 disassemble RNA granules in a post-translational modification-dependent manner, and NF45 targets other proteins in addition to NFAR2 (Fig. 9C).
Phosphorylation of NFAR2 at PKR sites induced a shift in its localization from the nucleus to the cytoplasm, which led to the enhancement of RNA granule assembly (Fig. 8, A-E). Although there are opposing results regarding the binding between phosphorylated NFAR2 and NF45, i.e. the current results showed that phosphorylated NFAR2 kept the binding to NF45 but a previous report showed that it did not (33), the net result was the same; phosphorylated NFAR2 increased its retention in the cytoplasm. Cytoplasmic retention of NFAR2 and NFAR2-TA was also increased by the expression of NF45 to almost the same level as that of NFAR2-TD, but they enhanced the assembly of RNA granules far less than NFAR2-TD (Fig. 8, F-H). These results suggested that phosphorylation of NFAR2 contributes not only to its cytoplasmic retention but also to overcoming RNA granule disassembly by NF45. Thus, when PKR is inactive, NFAR2 is localized in the nucleus and suppressed by NF45 even if a fraction of NFAR2 is exported to the cytoplasm. However, when PKR is activated, NFAR2 is phosphorylated, retained in the cytoplasm, and overcomes NF45 to enhance the assembly of RNA granules. The interplay between the positive and negative regulation of NFAR2 may function in the abrupt switching between the appearance and disappearance of RNA granules.
RNA granules (SGs) are induced by stress-responsive kinases, including PKR, which is activated by several kinds of stress such as viral dsRNA infection and peroxide or arsenite treatment (33,45,46). RNA granules contribute to translational arrest and cellular defense against stress. In this study, we showed that expression of NF45 disassembled RNA granules and rendered cells susceptible to arsenite stress, but arseniteinduced translational arrest was not affected (Fig. 3). This sustained translational arrest may be because NF45 could not affect the translational initiation pathway, including eIF2␣, although NF45 antagonizes translational repression by NFAR1/2 (Fig. 12). These results suggest that translational arrest is not sufficient for the protection of cells from stress and that RNA granule assembly has additional functions in cell survival, e.g. sequestration and nullification of stress-responsive apoptosis-promoting signaling molecules (53,54,10,11).
It was reported that knockdown of NF45 increases multinucleated cells, suggesting that NF45 is required for the proper progression of mitosis (42). Our present results that the association of NF45 with RNG105 mRNP complexes was increased in mitosis suggest a possible involvement of the regulation of RNA granules by NF45 in mitotic processes. Recently, RNA granule disassembly during meiosis was reported (55). RNA granules storing dormant cyclin B1 mRNA are formed in immature oocytes, but they are disassembled concomitantly with germinal vesicle (nuclear envelope) breakdown, which activates cyclin B1 mRNA translation. This RNA granule disassembly is reminiscent of our present result that RNG105-localizing RNA granules disappeared in mitosis. Whether NF45 is involved in the disassembly of RNA granules storing cyclin B1 mRNA and the progression of cyclin B1-dependent oocyte maturation and mitotic events are interesting issues to be solved in the future.
NFAR1 is ubiquitously expressed in several organs, but expression of NFAR2 is organ-specific (56). In particular, brain, testis, and skeletal muscle show high expression levels of NFAR2, which is coincident with the high levels of RNA granules in the brain and germ cells (3,5,57). Knock-out of NFAR1/2 leads to increased apoptosis, especially in cortical neurons and skeletal muscle (56). In skeletal muscle, NFAR1/2 bind to mRNAs for myogenic regulators such as MyoD and p21 WAF1/CIP1 and stabilize these mRNAs (56). It is notable that p21 WAF1/CIP1 mRNA also binds to the RNA granule component HuR (58,59) and is recruited into RNA granules to be stabilized (60). These reports suggest an association between NFAR2 function in RNA granules and skeletal muscle development and survival. Increased neuronal cell death and respiratory failure in NFAR1/2 knock-out mice suggest that NFAR1/2 also play important roles in neurons (56). Similar phenotypes have been reported in knock-out mice of other RNA granule components, including RNG105/caprin1 and G3BP (23,24). Elucidating the roles of NFAR1/2 in neuronal RNA granules as well as analyzing the genetic and functional interactions of NFAR1/2 with RNG105/caprin1, G3BP, and other RNA granule components in knock-out and mutant mice will be a topic for future studies.
NFAR2, particularly in its phosphorylated form, promoted large RNA granules. Although differences between large and small RNA granules are not well studied, large RNA granules are implicated in slower dissolution of mRNAs and other components from the granules (61, 62). Large RNA granules trigger eIF2␣ phosphorylation through PKR, suggesting a positive feedback regulation of granule assembly in large RNA granules (22). Large RNA granules are also induced by neurodegenerative disease-associated mutations in RNA granule components such as TDP-43, FUS/TLS, hnRNPA2B1, and hnRNPA1 (6 -8). These proteins contain prion-like LC sequence domains responsible for RNA granule assembly (6 -9), which has also been found in NFAR2. It is suggested that defective nuclear localization and cytoplasmic accumulation of disease-associated mutant TDP-43 and FUS are key events in neurodegenerative pathogenesis (6,7), which is reminiscent of our present result that increased cytoplasmic retention of phosphorylated NFAR2 was linked to the formation of large RNA granules. The structural and functional mechanisms of RNA granule assembly regulated by these RNA granule components provide a common basis for the generation of pathological large RNA granules in neurodegenerative disease.