SERPINE1 mRNA-binding protein 1 (SERBP1) is an arginine-methylated RNA-binding protein whose modification affects protein interaction and intracellular localization. In the present study, we show that, under normal growth conditions without stress, SERBP1 interacts with arginine-methylated and stress granule-associated proteins such as heterogeneous nuclear ribonucleoprotein A1, fragile X mental retardation protein and fragile X mental retardation syndrome-related protein 1 in an RNA-dependent manner. We also show that, after arsenite treatment, a proportion of full-length SERBP1 protein co-localizes with the typical stress granule marker T-cell intracellular antigen-1 in the cytoplasmic stress granules. Truncated SERBP1 with an N–terminal, central RG or C–terminal deletion, or single-domain segments comprising the N–terminal, central or C–terminal region, were recruited to stress granules upon arsenite treatment but with reduced efficiency. In addition, upon arsenite treatment, the localization of SERBP1 changed from a diffuse cytoplasmic localization to nuclear-dominant (concentrated in the nucleolus) A similar distribution was observed when cells were treated with the methylation inhibitor adenosine periodate, and was also detected for N- or C–terminal domain deletions and all three single-domain fragments even without stress induction. We further demonstrate that adenosine periodate treatment delays the association/dissociation of SERBP1 with stress granules. Hypomethylation retains SERBP1 in the nucleus/nucleolus regardless of arsenite treatment. Our study indicates that arginine methylation is correlated with recruitment of SERBP to stress granules and nucleoli and its retention therein. To our knowledge, this is the first report of an RNA-binding protein that is shifted simultaneously to cytoplasmic stress granules and nucleoli, two ribonucleoprotein-enriched subcellular compartments, upon stress.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
SERPINE1 mRNA-binding protein 1 (SERBP1) is an RNA-binding protein that was first named PAI-1 RNA binding protein 1 (PAI-RBP1) in recognition of its binding to the cyclic nucleotide-responsive sequence in plasminogen activator inhibitor (SERPINE) mRNA . It contains a central arginine and glycine (RG)-rich domain and a C–terminal arginine- and glycine-rich motif (RGG box) domain that are methylated by protein arginine N–methyltransferases (PRMTs). We have demonstrated protein arginine methylation in these regions of SERBP1, and the effect of methylation on the subcellular localization of SERBP1 . Hypomethylation by the treatment with the indirect methyltransferase inhibitor adenosine periodate (AdOx) or PRMT1 siRNA results in an increased nuclear distribution of SERBP compared with the otherwise predominant cytoplasmic localization. Analyzes of various deletions indicated that the central RG region is important for the nuclear localization, while both the N- and C–termini are required for nuclear export to the cytoplasm. Low methylation of the C–terminal RGG region also favors nuclear localization .
The Tudor domains of the survival of motor neuron protein and Tudor domain-containing protein 3 (TDRD3) are the best examples of specific methylarginine-binding modules . SERBP1 interacts with the Tudor domain of TDRD3 and co-localizes to cytoplasmic stress granules (SGs) . The SERBP1 protein was also reported to localize in SGs with the ORF1 protein of the LINE–1 retrotransposon . SGs are formed when cells face various stresses such as heat shock, oxidative stress or energy deprivation that halt translation initiation. SGs are dynamic structures containing messenger ribonucleoprotein particles (mRNPs) in which mRNAs have stalled at translation initiation, various translation initiation factors, and many RNA-binding proteins [6, 7]. SG formation is generally considered as a means of protection of stalled mRNPs for return to translation when the cells recover from the stress. They may be considered as a dynamic triage center, sorting mRNAs for storage, decay or re-initiation under stressful conditions. Another cytoplasmic RNP granule P–body appears to be involved in degradation of mRNA. However, some components are exchanged dynamically between the two [8, 9].
Many SG-incorporated proteins are RNA-binding proteins that shuttle between the nucleus and cytoplasm. Protein arginine methylation of the RG-rich motif of some RNA-binding proteins is critical for the subcellular distribution of the proteins, and may further affect their sorting to SGs . For example, mutations in the proline/tyrosine nuclear localization signal of fused in sarcoma (FUS) impair the normal nuclear localization of FUS. This may lead to the accumulated neuronal cytoplasmic inclusions of familial amyotrophic lateral sclerosis [11, 12]. Methylation inhibition restores nuclear import of the mutated FUS. Regulation of the methylation level of FUS alters its nuclear/cytoplasmic ratio . Part of the FUS in SGs co-localizes with PRMT1, and over-expressed PRMT1 reduces the proportion of FUS in insoluble inclusions . Fragile X mental retardation protein (FMRP) may trap mRNA in cytoplasmic granules and induce translation repression . Addition of AdOx increases the proportion of cells with granules, but decreases co-localization of FMRP granules with T-cell intracellular antigen-1 (TIA-1)-stained SGs . Ki–1/57, a paralog of SERBP1 with a similar protein arginine methylation pattern , interacts with FMRP and co-localizes to SGs, and appears to associate with ribosomes for translation regulation .
In this study, we found an association of SERBP1 with arginine-methylated and SG-associated proteins such as heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1), FMRP and fragile X mental retardation syndrome-related protein 1 (FXR1) through RNA. We also observed the presence of SERBP1 in SGs, and, unexpectedly, concentrated nucleolar localization under stress conditions with arsenite treatment. To our knowledge, this is the first report regarding relocation of an RNA-binding protein to these two RNP-enriched subcellular compartments upon stress.
Interaction of SERBP1 with other SG-incorporated RNA-binding proteins
SERBP1 has been reported to localize in cytoplasmic SGs that include other proteins such as the ORF1 protein of the LINE–1 retrotransposon  and the Tudor domain-containing protein TDRD3 . In our previous study, we observed the interaction of SERBP1 with asymmetric NG,NG–dimethylarginine-containing proteins by co-immunoprecipitation . As many SG-incorporated proteins are RNA-binding proteins and may be arginine-methylated , we analyzed whether SERBP1 interacts with these proteins. We first showed that hnRNPA1 co-immunoprecipitated with FLAG-tagged SERBP1 (Fig. 1A). The interaction requires RNA, because addition of RNase to the HeLa cell extract before immunoprecipitation totally blocked the interaction. However, interaction with PRMT1, which directly interacts with SERBP1 , was not influenced by RNase treatment. No hnRNPA1 co-immunoprecipitated from cell extracts prepared from cells transfected with empty vectors, confirming the specificity of the binding. Anti-FLAG was used to confirm immunoprecipitation (Fig. 1A, bottom panel). The reduced level of interaction with hnRNPA1 was not due to protein degradation by residual protease activities during RNase treatment because the protein level of hnRNPA1 in the extract was not affected by the reaction (Fig. 1B). Furthermore, we demonstrated co-immunoprecipitation of the asymmetric NG,NG–dimethylarginine-containing RNA-binding proteins FMRP, FXR1 and Ewing’s sarcoma protein (EWS) but not Src-associated in mitosis, 68 kDa (Sam68) with FLAG-tagged SERBP1. Interactions of these SG-associated proteins were also affected by RNase treatment (Fig. 1C). These results indicate that SERBP1 interacts with some RNA-binding proteins through RNA under normal conditions without stress. The central RG segment, which is important for SERBP1 nuclear localization , is not required or sufficient for the interaction, because RG deletion proteins interacted with hnRNPA1 and FMRP, but only very weak interaction was detected for the central segment (151–270) containing the RG sequences (Fig. S1). In comparison, SERBP1 constructs with N–terminal deletions (151–387) or the N–terminal domain alone (1–150) showed hnRNPA1 and FMRP interaction, indicating that this segment is not required but is sufficient for the interaction (Fig. S1).
Presence of SERBP1 in SGs upon arsenite treatment
The results presented above suggest a modest interaction of SERBP1 with other RNA-binding protein in the absence of stress. We then confirmed recruitment of SERBP1 to SGs upon stress. TIA–1, a typical cytoplasmic SG marker, was immunostained to determine the sites of SGs. In this study, we define TIA–1-positive cytoplasmic granules that occur after arsenite treatment as SGs. As shown in Fig. 2A, a small proportion of endogenous SERBP1 co-localized with TIA–1 at discrete cytoplasmic foci after arsenite treatment. We observed similar SG localization of GFP-tagged SERBP1 (Fig. 2B). There were no differences in SG formation between transfected and non-transfected cells. We observed co-localization of hnRNPA1 and SERBP1 in specific cytoplasmic foci (Fig. 2C), indicating that they may be recruited to the same SG.
Even though over-expression of cold inducible RNA-binding protein (CIRBP)  and FMRP  may promote SG assembly without further stress, we did not observe granule formation when SERBP1 was over-expressed by transfection. After arsenite treatment, no significant differences were detected for the number of SG-positive cells, SG number per cell or mean SG size between pEGFP-SERBP1- or pEGFP-transfected cells (data not shown).
We then evaluated which part of the protein is required for the SG localization. We constructed plasmids expressing GFP-tagged SERBP1 deletions or fragments (N–terminal, central RG or C–terminal; Figs 3A and S2A), and observed the expression patterns of the transfected cells. Before stress, full-length and RG-deleted SERBP1 proteins were present mostly in the cytoplasm and less so in the nucleus. SERBP1 proteins with a deleted N- or C–terminus predominantly localized in the nucleus before stress treatment (Fig. 3B; see also Fig. 5 of ). Upon arsenite challenge, the proportions of cells with SGs revealed by TIA–1 staining were similar whether the cells were transfected with full-length or truncated SERBP1 constructs (data not shown). Small proportions of SERBP1 deletions of the N–terminal, central RG or C–terminal domains migrate to SGs upon arsenite treatment, but the number of SERBP1-positive SGs is reduced. The reduction may be considered from a few perspectives. First, fewer successfully transfected cells contained SERBP1-positive SGs. The percentage of cells containing SERBP1-positive SGs decreased to ~ 70% for the central RG deletion and ~ 60% for the N- or the C–terminal deletions, compared to 80% for the full-length construct (Fig. 3 and Table 1). Second, fewer SGs (cytoplasmic foci stained with TIA–1) were SERBP1-positive (Table 1). The proportion of doubly SERBP1- and TIA–1-positive SGs decreased from 70% for the full-length protein to ~ 50% for the truncated ones. Furthermore, the number of SERBP1-positive SGs per cell and the SERBP1-positive area in SGs also decreased for the deletion proteins (Table S1). The results show that fewer SERBP1 deletion proteins are recruited to the SGs.
Table 1. Reduced SERBP1-positive SG formation for SERBP1 deletions or fragments.
151–387 (N-terminal deletion)
n is the number of GFP-positive cells counted.
n is the number of TIA–1-positive SGs counted. SGs were obtained from more than ten cells.
Proportion of cells with GFP–SERBP1-positive SGs (%)a
80.62 ± 1.41 (n = 120)
70.72 ± 3.85 (n = 89)
60.75 ± 0.39 (n = 102)
61.17 ± 3.49 (n = 114)
41.18 ± 3.05 (n = 71)
43.66 ± 2.82 (n = 64)
42.73 ± 2.31 (n = 103)
Proportions of TIA–1-positive SGs that are also SERBP1-positive (%)b
70.29 ± 2.52 (n = 346)
64.74 ± 2.37 (n = 242)
51.78 ± 0.85 (n = 232)
50.42 ± 1.11 (n = 260)
34.77 ± 6.34 (n = 204)
33.95 ± 0.83 (n = 194)
35.70 ± 3.18 (n = 192)
We then analyzed single-domain constructs comprising amino acids 1–150, 151–270 and 271–387). All three single-domain SERBP1 proteins re-localized to cytoplasmic SGs; however, even fewer cells contained SERBP1-positive SGs (~ 40%) and fewer SGs were SERBP1-positive (~ 35%; Fig. S2 and Table 1). The results indicate that any segment of SERBP1 is sufficient for SG localization, but complete SG distribution requires cooperation of more than one domain.
Increased nucleolar distribution upon arsenite treatment
Interestingly, we noticed that, although SERBP1 was primarily localized in the cytoplasm before any treatment, arsenite stress increased the nuclear distribution of SERBP. Similar patterns were detected by immunostaining of the endogenous SERBP1 protein using anti-SERBP1 antibodies (arrows in Fig. 2A) or by fluorescent microscopy for the GFP-tagged full-length protein (Fig. 2B). The nuclear SERBP1 was diffusely distributed throughout the nucleus but was concentrated in certain nuclear domains with no TIA–1 or 4′,6-diamidino-2-phenylindole (DAPI) staining. The nuclear distribution pattern was similar to that observed in our previous study upon addition of the indirect methyltransferase inhibitor AdOx (Figs 4 and 5 of ). Moreover, both the N- or C–terminal deletions, which mainly localized in the nucleus, were also concentrated in large nuclear bodies (Fig. 3; see also Fig. 5 of ). We observed that arsenite stress also induced strong nuclear distribution of the RG-deleted SERBP1 protein, which, like the full-length protein, was distributed predominantly in cytoplasm before treatment. Nuclear structures such as Cajal bodies, nuclear speckles and promyelocytic leukemia nuclear bodies are small foci (0.1–2.0 μm ), and thus are less likely to be the large nuclear bodies with intense SERBP1 staining that we observed. We speculated that the SERBP1-enriched large nuclear bodies are nucleoli by comparison of TIA–1 and DAPI staining patterns in various studies [4, 20]. The nucleolus distribution of SERBP1 was confirmed by co-localization of the SERBP1-enriched nuclear bodies (endogenous or GFP-tagged) with the nucleolus marker C23 (nucleolin; Fig. 4). The accumulation of SERBP1 was further quantified. Even though the total C23 and SERBP1 co-localized area was not significantly affected, the intensity of SERBP1 co-localized with C23 increased significantly after arsenite treatment (Table 2).
Table 2. Quantification of SERBP1 and C23 co-localization in nucleoli. The area and intensity of SERBP1 and C23 co-localization were defined using imagej software. Values are means ± SD from five cells. The intensity of SERBP1/C23 co-localization was significantly altered by arsenite treatment (P < 0.001).
1890.5 ± 345
38 328.2 ± 8208.2
2072.5 ± 781.8
106 748.1 ± 50 901.1
Methylation affects the recruitment of SERBP1 to SGs and retention therein
Inhibition of methylation appears to affect the assembly and disassembly dynamics of TDRD3-containing SGs . As SERBP1 interacts with TDRD3 and may be recruited to SGs, we determined whether the assembly/disassembly of SERBP1-containing granules is affected in a similar way. Treatment with the methylation inhibitor AdOx did not affect expression of endogenous or exogenous GFP-tagged SERBP1. The effect of methylation inhibition was confirmed by an in vitro methylation reaction. Proteins from AdOx-treated but not untreated HeLa cells were hypomethylated, and thus may be modified by the in vitro methylation reaction (Fig. 5A). After arsenite treatment, the percentage of cells containing TIA–1-positive SGs increased rapidly and reached ~ 90% after 30 min. The percentages increased at a similar rate whether the cells were pre-treated with AdOx or not. At recovery, the percentage decreased more slowly in the AdOx-treated cells at 120 and 150 min (Fig. 5B). The percentages of cells with SERBP1-enriched SGs (TIA–1- and SERBP1-positive) after 10 and 20 min of arsenite challenge were ~ 10% less than those with TIA–1-positive SGs (Fig. 5C). Nevertheless, the percentage of the cells with SERBP1-positive SGs reached ~ 90% after 30 min of the challenge. The proportion of cells with SERBP1-enriched granules decreased to ~ 50% after 90 min recovery. Only ~ 25% and 10% of SERBP1 granule-containing cells remained after recovery for 120 and 150 min, respectively. When the cells were pre-treated with AdOx and then challenged with arsenite, the proportion of cells with SERBP1-positive SGs was slightly less at early time points (10 and 20 min) but reached a similar level to that of untreated cells at 30 min. When arsenite stress was released, ~ 55% of cells had SERBP1-enriched SGs at 90 min. Most of these granule-containing cells remained at 120 and 150 min, leading to significant differences between AdOx-treated and untreated cells (Fig. 5C). In general, ~ 90% of the cells with TIA-positive SGs also contain SERBP1-positive SGs, indicating that SERBP1 is present in most but not all SGs. Without AdOx pre-treatment, only ~ 30–50% of the cells containing TIA-positive SGs are SERBP1-positive after recovery for 120–150 min. However, AdOx pre-treatment apparently retains SERBP1 in the SGs and greatly increases the ratio to ~ 100% after 150 min of recovery (Fig. 5D). The number of SGs per cell in cells pre-treated with AdOx increased more slowly and to a lesser extent than for untreated cells upon arsenite stress. After long recovery (120 and 150 min), the number of SGs per cell in AdOx-treated cells remained at a similar level (~ 12–15 per cell), while the number in untreated cells decreased gradually during recovery (from ~ 20 to fewer than ten per cell; Fig. S3). Overall, the results indicate that reduced methylation may hamper recruitment of SERBP1 to SGs but assist its retention.
Disappearance of SERBP1 from nuclei and nucleoli
We also examined the concentration of SERBP1 in nucleoli following addition and removal of AdOx or arsenite. Without treatment, ~ 20% of the cells showed a diffuse nuclear and concentrated nucleolar SERBP1 distribution. After treatment with arsenite or with arsenite and AdOx, the SERBP1 localization was nuclear-dominant in more than 90% of the cells. Compared with the disappearance of SERBP1-positive SGs after removal of the arsenite stress, the intense nuclear/nucleolar SERBP1 distribution of the SERBP1 protein disappeared more slowly. A high percentage (~ 60%) of the nucleus/nucleolus-dominant cells remained after 4 h, and slowly decreased to ~ 40% after 6 h of recovery (Fig. S4). The proportion of cells with a nuclear-dominant localization remained at ~ 40% after 12–24 h of stress release (data not shown). AdOx pre-treatment resulted in a slightly slower disappearance of SERBP1 from the nucleolus after removal of arsenite. Interestingly, slower exclusion of SERBP1 from the nucleus/nucleolus was observed for cells treated with AdOx only, compared with the cells treated with arsenite stress or both arsenite and AdOx (Fig. S4).
SGs form when translation initiation is blocked during cellular stress, in order to protect mRNA/mRNP for prompt re-entry into translation as soon as cells recover. Cytoplasmic SG formation is the standard response for cells facing arsenite challenge. In the present study, we first demonstrated the interaction of SERBP1 with SG-incorporated RNA-binding proteins such as hnRNPA1, EWS, FMRP and FXR1 by co-immunoprecipitation. Pre-treatment of the cell extracts with RNase abolished the interaction. SG localization and interaction of the LINE–1 ORF1p protein with SERBP1 and other RNA-binding proteins was observed. Association of ORF1p with RNA-binding proteins including hnRNPA1, DBP–A and YB–1 is also RNA-dependent . As the RNA-dependent interaction of SERBP1 and other RNA-binding proteins occurs in the absence of stress, whether the interaction generates further granule assembly under stress conditions is an interesting issue. Intact RNA may be the critical scaffold for concentration and aggregation of interacting RNA-binding proteins to SGs under stress conditions. We specifically observed co-localization of SERBP1 and hnRNPA1 in SGs upon arsenite treatment.
In this study, we showed that most typical SGs stained by TIA–1 co-localized with SERBP1-enriched foci after arsenite challenge. SERBP1 with the N–terminal, central RG or C–terminal domain deletions still localized to SGs but in a smaller proportion of cells and in fewer SGs. Further dividing SERBP1 into N–terminal, central and C–terminal segments showed that all three single domains of SERBP1 localize to cytoplasmic SGs, but with lower efficiencies. The results indicate that, even though any single segment of SERBP1 is sufficient for SG localization, the three domains collaborate for complete SG distribution. However, the predominantly nuclear-distributed N- or C–terminal deletion proteins and the single-domain fragments of SERBP1 must move to the cytoplasm to form SGs upon arsenite treatment. Alternatively, residual cytoplasmic SERBP1 concentrated in the granules under stress.
We have shown that SERBP1 is arginine-methylated in the central RG and C–terminal RGG regions . Protein arginine methylation of some SG-incorporated RNA-binding proteins has been shown to be critical for their cellular distribution, and may further affect their sorting to SGs . We are interested in whether arginine methylation of SERBP1 is related to its SG localization. We did not observe effects of AdOx treatment on interaction of SERBP1 with other SG-incorporated RNA-binding proteins such as hnRNPA1 or FMRP, or on localization of SERBP1 to SGs (data not shown). Deletion of either the central RG or C–terminal RGG box did not block the SG redistribution, indicating that arginine methylation of SERBP1 may not affect the response. However, we did observe slow recruitment of SERBP1 and reduced disassembly of SERBP1-enriched SGs after AdOx treatment, similar to what was observed for TDRD3-enriched SGs . Although it is still possible that reduced arginine methylation affects the general mechanism of SG formation, component selection and/or SG disassembly, SERBP1-positive SGs respond to AdOx treatment more significantly than total SGs (stained by TIA–1). It is thus likely that arginine methylation of SERBP1 affects the association/dissociation of SERBP1 to SGs but not the assembly and disassembly of SGs. Determination of whether hypomethylated SERBP1 is less dynamic than the normally methylated SERBP1 requires further experimentation.
Most interestingly, we found that, under arsenite treatment, full-length as well as RG-deleted SERBP1 had a tendency to migrate to the nucleus, and was concentrated in the nucleolus in addition to the obvious cytoplasmic SG formation. The site for intense nuclear distribution of SERBP1 was determined as the nucleolus by co-localization with the nucleolar marker nucleolin. Nucleoli are sub-nuclear compartments in which ribosome biogenesis takes place, and accumulate factors involved in rRNA transcription, processing and ribosome subunit assembly. Over recent years, diverse functions of the nucleolus related to sequestration and release of specific proteins for regulation of transcription regulation, protein–protein interaction and growth control have been suggested . The nucleolar dynamics under stress responses have been discussed , but not together in conjunction with cytoplasmic SG formation. We observed nucleolar accumulation of SERBP1 upon arsenite treatment in this study. A similar localization was observed with deletions of the N- or C–terminus, and with all single-domain fragments of SERBP1 even without stress. A stress granule and nucleolar protein implicated in ribosomal processing is localized in the nucleolus and cytoplasm. Upon arsenite exposure, the stress granule and nucleolar protein was restricted to cytoplasmic SGs with reduced nucleolar staining . There have been no reports of simultaneous accumulation of an individual protein in cytoplasmic SGs and nucleoli in response to arsenite challenge. To our knowledge, this is the first report to describe recruitment of a protein to both cytosolic SGs and nucleoli after arsenite challenge, and may indicate possible coordination or similar assembly/disassembly mechanisms.
Evidence for the interactions of SERBP1 with ribosomal proteins has been obtained in various proteomic studies. A search of the online interaction repository biogrid (Biological General Repository for Interaction Datasets, version 3.2.98; http://thebiogrid.org/), which comprises 38 590 publications describing 648 491 raw protein and genetic interactions, showed 61 interactors for SERBP1. Among these, 12 are ribosomal proteins. A further interactome study revealed that SERBP1 co-elutes with more than 40 polypeptides, mostly ribosomal 40S proteins or subunits of eukaryotic initiation factor 3 . The interaction evidence raises the possibility that stresses that block translation initiation may lead to accumulation of ribosomal proteins/subunits in the nucleoli and trap the interacting SERBP1. The stress condition or SERBP1 deletions may strengthen the association of SERBP1 with ribosome proteins, and result in accumulation of the protein in the nucleolus. Nevertheless, SERBP1 is not included in the NOPdb (Nucleolar Proteome Database), which includes more than 4500 human proteins identified by mass spectrometry from purified nucleoli . No nucleolar distribution of SERBP1 was suggested by nucleolar proteomic analyzes of the dynamic distribution under virus infection [25, 26] or DNA damage . However, two nucleolar localization signals in the N–terminus were determined using NoD . Another putative signal with a value slightly below the 0.8 threshold was present at the C–terminus (Fig. 6). Whether these elements explain the nucleolar localization of SERBP1 deletions or fragments requires further investigation.
Recently, cytosolic or nuclear non-membrane-bound compartments such as SGs, as well as nucleoli, have been described as a type of large aggregate with a pool of dynamic soluble components. They often contain various ribonucleoproteins comprising both protein and RNA . It was suggested that, for a protein to aggregate in RNA granules, both the RNA-binding domain and low-complexity sequences with little sequence diversity must be present . The vasa intronic gene, the SERBP1 ortholog in Drosophila, has been identified in a screen for RNA granule formation . The structures of low-complexity sequences are close to the cross β–structures typical of the fibrils in neurodegenerative Alzheimer's or prion disease . Pathogenic mutations of RNA-binding proteins such as hnRNPA1 have been shown to strengthen fibril formation and may lead to multi-system proteinopathy . SG proteins such as TIA–1 and TIA–R have been reported to contain QN-rich prion-like domains, and over-expression of these proteins promotes self-aggregation and SG assembly . Over-expression of CIRBP1  and FMRP  also promotes SG assembly. However, we did not observe granule formation as a result of over-expression of SERBP1. However, putative fibril-forming segments within SERBP1 are predicted by ZipperDB (http://services.mbi.ucla.edu/zipperdb/) , a method that predicts the tendency to form self-complementary β–strands within proteins. Segments with scores below the −23 kcal·mol threshold for high fibrillation propensity are present in the N–terminus as well as the central domain and C–terminus of SERBP1 (Fig. 6). Thus, the three segments of SERBP1 each contain a ‘steric zipper’ and an RNA-binding region. This may explain why we observed distribution of all three fragments or deletions to SGs.
We had shown previously that inhibition of methylation by AdOx or PRMT1 siRNA shifts the localization of full-length and central RG-deleted SERBP1 towards the nucleus . In this study, we showed that arsenite treatment resulted in redistribution of SERBP1 to SGs and the nucleolus simultaneously. AdOx treatment alone, like another methylation inhibitor methylthioadenosine (MTA), did not lead to SG formation  but only predominant nuclear/nucleolar distribution of SERBP1. Therefore, the mechanisms and critical factors for distribution of SERBP1 to the two non-membrane-bound compartments appear to be different. The association and dissociation of SERBP1 to SGs are slower in hypomethylated cells. Likewise, we observed the difference in relocation of SERBP1 in the nucleolus due to methylation. Hypomethylated SERBP1 remained in the nucleus/nucleolus regardless of treatment with arsenite or not. It is not known whether the induction of nucleus/nucleolus distribution of SERBP1 by arsenite and hypomethylation occurs through similar pathways, but the effects of arsenite and methylation inhibition on retention of SERBP1 in the nucleus/nucleolus were not additive.
In conclusion, our study showed that, in the absence of stress, SERBP1 binds to various RNA-binding proteins that are also present in SGs. The interaction is RNA-dependent, and may imply further assembly of these proteins in SGs as concentrated aggregations of RNPs. The cytoplasmic SG and nucleolar localization after arsenite treatment observed in the present study is the first demonstration of relocation of an RNA-binding protein to these two RNP-enriched subcellular compartments upon stress. We showed that hypomethylation may lead to slower movement of SERBP1, reflected by decreased dissociation from SGs or longer retention in the nucleolus.
The SERBP1 constructs were prepared as described previously . SERBP1 fragments to be fused with the GFP tag were generated by PCR using primers 5′-AAGAATTCCATGCCTGGGCACTTACAGG-3′ and 5′-AAGGATCCTTATTCGCCTCCTTCACCC-3′ (N–terminal domain, amino acids 1–150), 5′-GCGAATTCATTTTCAGTTGATAGACCG-3′ and 5′-GCGGATCCTTAATCCAAAGTCATCTC-3′ (central RG domain, amino acids 151–270), and 5′-GCGAATCCAGAGTGGAAGGCTATTCAA-3′ and 5′-AAGGATCCAGTTAAGCCAGAGCTGG-3′ (C–terminal domain, amino acids 271–387). The resulting PCR products were inserted into the TA vector (Yeastern Biotech, Taipei, Taiwan), and then sub-cloned into pEGFP (Clontech Laboratories Inc., Shiga, Japan) via the EcoRI/BamHI sites.
Cell culture and treatments
The human HeLa cervical carcinoma cell line (American Type Culture Collection stock number CCL–2) was maintained at 37 °C in minimal essential medium (Gibco/Life Technologies, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (BioWest, Nuaillé, France), 100 U·mL penicillin, 100 μg·mL streptomycin and 2 mm l–glutamine (Thermo, Waltham, MA, USA) in a 5% CO2 incubator. Cells in a 35 mm dish grown to ~ 50% confluence were transfected with plasmids using Lipofectamine 2000 (Invitrogen/Life Technologies, Grand Island, NY, USA) according to the manufacturer's instructions. After 24 h of transfection, the cells were treated with 1 mm sodium arsenite (Sigma-Aldrich, St. Louis, MO, USA) for 1 h, followed by 30 min recovery in normal medium to induce a cellular stress response. To inhibit methytransferases, transfected HeLa cells were incubated in normal minimal essential medium containing 50 μm adenosine-2′,3′-dialdehyde (AdOx; Sigma-Aldrich) for 24 h, and stress was subsequently induced by addition of 0.5 mm arsenite for 30 min.
Untreated and treated HeLa cells were cultured on glass cover slips and fixed with 2% paraformaldehyde (Sigma-Aldrich) in NaCl/Pi at room temperature for 15 min. After washing three times for 5 min with NaCl/Pi at room temperature, the fixed cells were permeabilized for 5 min at room temperature with NaCl/Pi containing 0.5% Triton X–100 and washed again as described above. Blocking was performed using NaCl/Pi containing 1% bovine serum albumin (BSA/NaCl/Pi) for 60 min at room temperature. The cells were incubated with primary antibodies at a 1 : 200 dilution for anti-SERBP1 mouse monoclonal antibodies from Abnova (Taipei City, Taiwan) and a 1 : 500 dilution for anti-SERBP1 rabbit polyclonal antibody [prepared in our laboratory by immunizing rabbits with GST-SERBP1 (151–270)]; 1 : 200 dilution for anti-nucleolin (C23) or anti-TIA-1 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 4 °C overnight, washed three times for 5 min with NaCl/Pi containing 0.1% Tween–20, and incubated in NaCl/Pi containing 0.1% Tween–20 and fluorescein isothiocyanate-conjugated anti-mouse, anti-goat antibody, rhodamine-conjugated anti-rabbit or anti-goat antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) for 1 h, followed by DAPI (0.5 μg·mL; Roche, Mannheim, Germany) for 5 min in the dark at room temperature. The cells were washed three times with NaCl/Pi containing 0.1% Tween–20, and observed using a fluorescent microscope (Ziess, Oberkochen, Germany). Protein co-localization analysis was performed by counting singly and doubly stained granules of representative images using imagej (http://rsb.info.nih.gov/ij/).
Immunoprecipitation and immunoblotting
After 48 h of transfection, the cells were harvested and cell extracts were prepared as described previously . For RNase A treatment, equal portions of FLAG plasmid-transfected HeLa cell extracts were incubated with RNase A (Sigma-Aldrich) at room temperature for 30 min prior to clarification by centrifugation (12 000 g, 10 min, 4 °C). RNase-treated and untreated extracts were incubated with anti-FLAG M2 affinity gel (Sigma-Aldrich) at 4 °C overnight. The supernatant was removed by centrifugation (8200 g, 3 min, 4 °C), and the beads were washed three times with Tris-buffered saline (50 mm Tris/HCl pH 7.5, 150 mm NaCl, 1 mm EDTA). The FLAG-tagged proteins were eluted using FLAG elution buffer containing 3x FLAG peptide (Sigma-Aldrich). Total HeLa cell extract or the FLAG-tagged proteins were separated by SDS/PAGE, and subsequently transferred to nitrocellulose membranes (Sartorius Stedim Biotech, Aubagne, France). The membranes were blocked in 7% skimmed dry milk in Tris-buffered saline containing Tween–20 (10 mm Tris/HCl pH 7.5, 100 mm NaCl, 0.1% Tween 20) for 1 h, incubated overnight with primary antibodies [1 : 2000 dilution for anti-FMRP, 1 : 2500 dilution for anti-PRMT1, 1 : 2000 dilution for anti-PRMT5 (all from Merck/Millipore, Darmstadt, Germany), 1 : 1000 dilution for anti-FXR2 (BD Biosciences, San Jose, CA, USA), 1 : 1000 dilution for anti-EWS, anti-TIA–1 and anti-hnRNPA1 (all from Santa Cruz Biotechnology), and 1 : 5000 dilution for anti-FLAG (Sigma-Aldrich)], washed three times for 5 min in Tris-buffered saline containing Tween–20, then incubated with secondary antibody (horseradish peroxidase-conjugated anti-mouse, -rabbit or -goat IgG) for 1 h, and washed again as described above. Chemiluminescent detection was performed using the VisGlow substrate for horseradish peroxidase (Visual Protein, Taipei, Taiwan) according to the manufacturer's instructions.
All data are the results of at least three independent experiments, and were statistically analyzed using microsoft excel (Microsoft company, Redmond, WA, USA). Values are means ± standard errors of the mean (SEM). Statistical significance was determined using Student's t test.
The project was supported by grants 98–2320–B–040–011–MY3 and 101–2320–B–040–004 from the Taiwan National Science Council, and grant number 98–OM–A–060 from Chung Shan Medical University. Fluorescence microscopy was performed in the Instrument Center of Chung Shan Medical University, which is supported by the Taiwan National Science Council, Taiwan Ministry of Education and Chung Shan Medical University.