Research Article 981 The translational regulator CPEB1 provides a link between dcp1 bodies and stress granules

A. Wilczynska*, C. Aigueperse, M. Kress, F. Dautry and D. Weil‡ CNRS UPR1983, Institut André Lwoff, 7 rue Guy Moquet, 94801 Villejuif CEDEX, France *Present address: Molecular Biology Department, Cancer Center-Institute, ul. Roentgena 5, 02-781 Warsaw, Poland ‡Author for correspondence (e-mail: [email protected])

Accepted 21 December 2004 Journal of Cell Science 118, 981-992 Published by The Company of Biologists 2005 doi:10.1242/jcs.01692

Summary The cytoplasmic polyadenylation element-binding protein clam, and as a component of dcp1 bodies in mammals, is (CPEB) has been characterized in Xenopus laevis as a also present in stress granules. Both stress granules and translational regulator. During the early development, it dcp1 bodies are involved in mRNA storage and/or behaves first as an inhibitor and later as an activator of degradation, although so far no link has been made translation. In mammals, its closest homologue is CPEB1 between the two, in terms of neither morphology nor for which two isoforms, short and long, have been protein content. Here we show that transient CPEB1 described. Here we describe an additional isoform with a expression induces the assembly of stress granules, which different RNA recognition motif, which is differentially in turn recruit dcp1 bodies. This dynamic connection expressed in the brain and ovary. We show that all CPEB1 between the two structures sheds new light on the isoforms are found associated with two previously compartmentalization of mRNA metabolism in the described cytoplasmic structures, stress granules and dcp1 cytoplasm. bodies. This association requires the RNA binding ability of the protein, whereas the Aurora A phosphorylation site is dispensable. Interestingly, the rck/p54 DEAD box Key words: CPEB, dcp1 body, GW body, Translation, Storage, protein, which is known as a CPEB partner in Xenopus and Degradation

Introduction oocytes and BDNF (brain derived neurotrophic factor) and

Journal of Cell Science The translational control of mRNA is a way of regulating the NMDA (N-methyl-D-aspartate) in neurons, all of them leading production of proteins from pre-existing mRNAs within a short to the activation of the Aurora A kinase. This kinase then time scale. In higher eukaryotes, a well-known example is the directly phosphorylates CPEB protein (Wells et al., 2000). In regulation by cytoplasmic polyadenylation, which has been the Xenopus oocyte, dephosphorylated CPEB is bound to mainly described in Xenopus during oocyte activation and maskin (Stebbins-Boaz et al., 1999). Maskin interacts with the early development. As the mature oocytes have neither a translation initiation factor eIF4E present at the 5′ extremity of transcriptional nor an mRNA degradation activity, the the target mRNA, preventing the recruitment of eIF4G, and regulation of gene expression relies solely on this translational thus translation initiation. Following phosphorylation of control. During oogenesis many mRNAs are hypoadenylated CPEB, maskin releases eIF4E, whereas CPEB recruits the and stored in a dormant state. Following oocyte activation, they polyadenylation factor CPSF (cleavage and polyadenylation are readenylated under the control of the cytoplasmic specificity factor) to the polyadenylation signal, leading to the polyadenylation element-binding protein (CPEB) and become subsequent polyadenylation of the mRNA (Mendez et al., competent for translation. Most of the known CPEB target 2000). This finally results in the resumption of translation mRNAs encode proteins that play a direct role in meiosis and initiation. mitosis, such as the mos kinase, cyclins and the Eg5 kinesin. What is the fate of untranslated mRNAs within the cell? This regulation is essential for the resumption of meiosis and They can be stored in specific sites awaiting a signal able to the early embryonic development (Groisman et al., 2000). trigger their translation. This is for instance the case of the These observations have been extended to mammalian neuronal mRNAs localized at synapses, which are translated development and somatic cells, in particular to neurons, where following synapse activation. A well-documented example, α- CPEB target mRNAs are translated following synapse CaMKII mRNA, is in fact a CPEB target (Aakalu et al., 2001; activation (Wells et al., 2000). In this case, the translational Wu et al., 1998). Alternatively, they can be routed to regulation is one of several control levels, as somatic cells cytoplasmic structures that will direct them towards storage or continuously transcribe and degrade mRNAs. It may enable a degradation pathways. Two such structures have been rapid and local response of the cells to the stimulation of a described in fibroblasts: stress granules and dcp1 bodies. single synapse. At the molecular level, CPEB activity is In response to environmental stress such as oxidative regulated by external stimuli such as progesterone in Xenopus conditions, UV irradiation or heat shock, global translation 982 Journal of Cell Science 118 (5)

diminishes. In particular, the translational initiation factor (accession number H12139/IMAGE:48054). The CPEB1-long and eIF2-α is phosphorylated and therefore unable to reload GTP -short plasmids were derived from the previous plasmids by restriction and recruit eIF5 to the preinitiation complex. As a result, fragment exchange with the non-∆5 cDNA (BG722145/IMAGE: abortive eIF2/eIF5-deficient 48S* complexes assemble on 4830450). F314A, H545A, T172D and T172A point mutations were mRNAs and accumulate in cytoplasmic granules together with created in the CPEB1-long expression vector using the QuickChange various mRNA binding proteins (Anderson and Kedersha, XL site-directed mutagenesis kit (Stratagene, The Netherlands). The Tet-regulated vector has been derived previously from the tTA 2002). These so-called stress granules are thought to form as expression vector (Dirks et al., 1994; Weil et al., 2000). The a result of intermingling of blocked mRNAs through the self- tTA/CPEB1-∆5-long plasmid was created by inserting a restriction aggregation of the RNA binding proteins TIA-1 and TIAR. In fragment containing the full ORF of the pEGFP/CPEB1-∆5-long the absence of stress, the overexpression of a phosphomimetic plasmid. The red fluorescent protein (RFP) cDNA was inserted mutant of eIF2-α leads to the assembly of similar stress upstream of the human rck/p54 cDNA in pCMV-SPORT6 vector granules (Kedersha et al., 2002). The fact that stress granules (BC065007/IMAGE:6163439), so that the complete p54 ORF is in- are dynamic and reversible structures suggests that they could frame with the RFP ORF at the N-terminus. be sorting sites for the storage or degradation of untranslated mRNAs (Kedersha et al., 2002). Cell culture The machinery of the main mRNA degradation pathway, the ′ ′ Epitheloid carcinoma HeLa cells and retinal pigment epithelial RPE- 5 -3 mRNA decay, is concentrated in cytoplasmic foci both in 1 cells (BD Biosciences Clontech, France) were routinely maintained mammals (Ingelfinger et al., 2002; van Dijk et al., 2002) and in DMEM and DMEM/F12, respectively, supplemented with 10% in yeast (Sheth and Parker, 2003), called dcp1 bodies and P- fetal calf serum. For stress induction, cells were treated with 0.5 mM bodies, respectively. This includes the decapping enzymes arsenite (Sigma Aldrich, France) for 30 minutes, followed by recovery dcp1 and dcp2, the 5′-3′ exonuclease Xrn-1, proteins that for 30 minutes in the absence of arsenite (Kedersha et al., 2002). stimulate decapping such as the LSm complex and the Translation inhibition was achieved using 10 µg/ml cycloheximide Dhh1p/rck/p54 protein, as well as an RNA binding protein of (Roche Diagnostics, France) for 40 minutes or 100 µg/ml puromycin unknown function, GW182, initially identified as a human (Sigma) for 1 hour. µ autoantigen (Eystathioy et al., 2002). Blocking the mRNA Transient transfections were performed with 1.5 g plasmid DNA degradation pathway by genetic means in yeast (Sheth and per 35 mm diameter dish by a standard calcium phosphate procedure (Sambrook et al., 1989). For stable transfection, HeLa rtTA HR5 (BD Parker, 2003) or using siRNAs in mammals (Cougot et al., Biosciences Clontech, France) cells were co-transfected with 6 µg 2004) enhanced the accumulation of mRNAs in these foci, tTA/CPEB1-∆5-long plasmid and 0.6 µg of hygromycin-resistant pY3 indicating that they are active sites of mRNA degradation plasmid DNA (Blochlinger and Diggelmann, 1984) per 100 mm rather than storage sites for these factors. Thus, mRNA diameter dish using lipofectamine 2000 (Invitrogen, France). The degradation, at least in part, occurs in specific bodies and not selection of stable clones was achieved as described previously diffusely throughout the cytoplasm as postulated previously. (Audibert et al., 2002). Cells were then routinely maintained in the Although one CPEB gene has been described in Xenopus presence of 100 µg/ml geneticin sulfate (Invitrogen, France) and 200 laevis, four genes have been reported so far in mammals, µg/ml hygromycin (Invitrogen, France). Induction of the Tet promoter µ named CPEB1 to CPEB4 (Gebauer and Richter, 1996; was performed by addition of 1 g/ml doxycycline to the culture Journal of Cell Science Kurihara et al., 2003; Theis et al., 2003; Welk et al., 2001), medium. CPEB1 clearly being the closest homologue to the Xenopus gene. In addition, two isoforms, long and short, resulting from RT-PCR differential exons at the 5′ extremity of the mRNA, have been Reverse transcription reactions were performed with 1 µg total RNA described in humans (Welk et al., 2001). We have characterized using random primers and Mu-MLV reverse transcriptase (Invitrogen, a new CPEB1 isoform that differs in its first RNA recognition France). One tenth of the reaction was used for PCR amplification. domain. Both previously described long and short isoforms as The position of the CPEB1 primers is indicated in Fig. 1A. The rrm5 well as this new isoform are differentially expressed in the (5′-ATGGCCAGGAGCTTCTGT-3′) and rrm3 (5′-CGGCTGGA- ovary and brain, two tissues where CPEB activity has been CATCTTGAAA-3′) primers were used for the amplification of all CPEB1 isoforms, lg5 (5′-GGGGTACCGCTGGGACAACCAAG- demonstrated (Groisman et al., 2000; Huang et al., 2002). All ′ ′ ′ CPEB1 isoforms were recruited to two structures involved in GAAG-3 ) and ls3 (5 -GACTAGTCCAAGTCAGACCCAAGGG-3 ) for the amplification of the CPEB1-long isoform and sh5 (5′- mRNA storage and/or degradation: stress granules and dcp1 GGGGTACCAGCGGGAAGCATCAGCAG-3′) and ls3 for the am- bodies. In addition, GFP-fused CPEB1 protein was able to plification of the CPEB1-short isoform. After 37 cycles (94°C for 45 assemble stress granules in the absence of stress. Finally, these seconds, 60°C for 45 seconds and 72°C for 45 seconds), amplification granules in turn recruited components of dcp1 bodies, products were separated on a 1.2% agarose gel and stained with indicating a dynamic relationship between these two structures. ethidium bromide. Each primer set targeted two separate exons so that the amplification could be unambiguously attributed to reverse- transcribed mRNAs rather than residual genomic DNA. To distinguish Materials and Methods between the ∆5 and non-∆5 isoforms, the rrm5-rrm3 amplification Expression vectors products were digested with EcoNI restriction enzyme, separated on The human CPEB1-∆5-long cDNA sequence from nucleotides 1 to an 8% polyacrylamide gel and stained with ethidium bromide. 1791 (GenBank accession number BC050629/IMAGE:6047179) was inserted into the eukaryotic expression vector pEGFP-N1 (BD Biosciences Clontech, France), so that the complete CPEB1-∆5-long Immunofluorescence open reading frame (ORF) was in-frame with the GFP ORF at the C- To raise anti-human CPEB1 antibodies, the human CPEB1-∆5 cDNA terminus. The CPEB1-∆5-short expression vector was derived from sequence from nucleotides 310 to 2090 (GenBank accession number this vector by restriction fragment exchange with the short cDNA BC050629/IMAGE:6047179) was inserted into the prokaryotic dcp1 bodies and stress granules 983

Fig. 1. Alternative splicing of CPEB1 within the RRM1. (A) Schematic representation of hCPEB1 mRNAs. Human CPEB1-long (lg) and -short (sh) mRNAs are represented with the ORF in black including the two RRM domains in grey and the zinc finger domain (Zn). The short specific 5′ UTR is hatched, with the optional intron indicated with a star. The primers and the EcoNI restriction site used are indicated by arrows and scissors, respectively. The black triangle indicates the position of the alternative 15 nucleotides. (B) Comparison of CPEB1 of various species. Partial RRM1 sequences from human, murine, Xenopus and Zebrafish EST and cDNA were aligned. The GT dinucleotide corresponding to a splice donor site is highlighted in grey. The encoded amino acids are indicated below the sequence. hsCPEB1-∆5 is from GenBank accession number BX327041 (nucleotides 586-630), hsCPEB1 from AF329402 (nt. 1355-1414), mmCPEB1-∆5 from BI144277 (nt. 395-439), mmCPEB1 from NM_007755 (nt. 1064-1123), xlCPEB from XLU14169 (nt. 1114-1173) and drCPEB from AF076918 (nt. 1032-1091). (C) Position of the deletion with respect to RRM structure. α helices and β sheets of the RRM are represented, the black triangle indicating the position of the alternative five amino acids. (D) Differential expression of CPEB1-long and -short in tissues and cell lines. CPEB1 mRNA from indicated samples was amplified by RT-PCR using common primers rrm5 and rrm3 (upper panel), CPEB1-long specific primers lg5 and ls3 (middle panel) or CPEB1-short specific primers sh5 and ls3 (lower panel), as illustrated in A. Amplification in the absence of RNA was used as a negative control (–). A 100 bp ladder was used as molecular weight marker (M). (E) Differential expression of the ∆5 isoform in tissues and cell lines. RNA was amplified by RT-PCR using rrm5 and rrm3 primers and digested with EcoN1, as illustrated in A. The ψX174 HaeIII digest was used as molecular weight marker (M).

expression vector pGEX6P3 (Amersham Pharmacia, France), so that HCl, pH 8, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40) the CPEB1 ORF was in-frame with the GST ORF at the N-terminus. supplemented with complete protease inhibitor cocktail (Roche This plasmid was transformed into Escherichia coli BLR (DE3) Diagnostics, France) and incubated on ice for 30 minutes. Soluble bacteria. The GST-tagged CPEB1 protein was purified on a proteins were recovered after centrifugation at 15,000 g at 4°C for 10 preparative SDS-PAGE gel and injected into mice. Monoclonal minutes and quantified by the Bradford method (BioRad, France). antibodies recognizing the CPEB1 moiety of the fusion protein by Proteins were separated on a 7.5% polyacrylamide SDS-PAGE gel Journal of Cell Science immunofluorescence and western blotting were selected. along with a prestained protein ladder 10-180 kDa (MBI Fermentas, The anti-GW182 human index serum was a kind gift from France) and transferred to a nitrocellulose membrane (Amersham Theophany Eystathoy (University of Calgary, Alberta, Canada), the Pharmacia, France). Non-specific protein binding sites were blocked anti-hDcp1 rabbit antibody from Bertrand Séraphin (Centre de by incubation in PBST (PBS with 0.1% Tween-20) containing 5% Génétique Moléculaire, Gif, France) and the anti-eIF3 goat antibody (w/v) non-fat dry milk for 1 hour at room temperature. The from John Hershey (University of California, Davis, CA). The membrane was then incubated with the monoclonal antibody (1:5000 secondary antibodies conjugated to rhodamine and FITC were in PBST containing 5% non-fat dry milk) overnight at 4°C. After purchased from Jackson Immunoresearch Laboratories (Immunotech, washing in PBST, the blot was incubated with horseradish France). peroxidase-conjugated goat anti-mouse IgG (1:10,000) (Immunotech For immunofluorescence, cells were grown on glass coverslips and Jackson, France) for 45 minutes at room temperature. After washing either fixed with 4% paraformaldehyde for 10 minutes and in PBST, immune complexes were detected using the SuperSignal® permeabilized with 0.5% Triton X-100, 50 mM NH4Cl for 15 minutes chemiluminescent substrate detection reagent (Perbio Science, for anti-hDcp1 staining, or fixed in –20°C methanol for 3 minutes for France). anti-GW182 and anti-eIF3 staining. Cells were incubated with the primary antibody for 1 hour, rinsed with phosphate-buffered saline (PBS), incubated with the secondary antibody for 30 minutes, rinsed Results µ with PBS and stained with 0.12 g/ml DAPI for 1 minute, all steps Variable CPEB1 isoforms in mammalian cells being performed at room temperature. Slides were mounted in Citifluor (Citifluor, UK) and observed on a Leica DMR microscope Although only one CPEB isoform has been described in (Leica, Heidelberg, Germany) using a ×63/1.32 oil-immersion Xenopus laevis, two human CPEB1 mRNAs have been objective. Photographs were taken using a Micromax CCD camera reported, encoding a long and a short CPEB1 isoform (Welk (Princeton Instruments). Confocal images were obtained on a Leica et al., 2001). The long form is 566 amino acids long and very TCS-NT/SP inverted confocal laser-scanning microscope using an similar to the Xenopus CPEB protein. The short form Apochromat ×63/1.32 oil-immersion objective. Fluorescence signals corresponds to a 68 amino acid N-terminal truncation of the were acquired in 0.16 µm optical sections. long form, owing to the presence of alternative upstream exons (Fig. 1A). Western blot analysis We have identified in expressed sequence tag (EST) Cells were scraped into PBS, resuspended in lysis buffer (50 mM Tris- databanks a new CPEB1 mRNA containing a 15 nucleotide 984 Journal of Cell Science 118 (5) deletion in the open reading frame both in the human and mouse (Fig. 1B). Based on the human genome sequence, these 15 nucleotides are located precisely at the end of an exon. The deletion starts with a consensus GT dinucleotide, confirming that it results from an alternative splicing donor site. This variant mRNA encodes a CPEB1 protein with a five amino acid deletion (GNMPK) located in the first RNA recognition motif (RRM1) upstream of the third β-sheet (Fig. 1C). We named it CPEB1-∆5. The expression of the long and short mRNAs and the presence of the ∆5 deletion were investigated by RT-PCR in murine brain and ovary tissues, as well as in the two human cell lines HeLa and RPE-1. The CPEB1 long form mRNA was present in the four samples, whereas the short form was detected only in the brain (Fig. 1D). In addition, the short form Fig. 2. CPEB1 localization in HeLa cells. (A) Western blot assay. was amplified as two bands, probably corresponding to an The anti-CPEB1 monoclonal antibody 2B7 was used for western optional intron of 211 nucleotides, which is present within the blotting of proteins from HeLa cells expressing the murine CPEB1- ∆5-short or the human CPEB1-∆5-long isoform. Untransfected HeLa 5′ UTR in several murine and human ESTs. To distinguish ∆ ∆ cells were used as a control. (B) Immunofluorescence of between the 5 and non- 5 isoforms, the region encompassing untransfected HeLa cells. Cells were fixed, stained with anti-CPEB1 the alternative 15 nucleotides was amplified by RT-PCR, antibody and observed by fluorescence microscopy. Bar, 8 µm. cleaved by the EcoNI restriction enzyme in order to reduce its size (Fig. 1A) and analysed on a polyacrylamide gel (Fig. 1E). Although both forms could be amplified from all samples, the non-∆5 form was most abundant in the ovary, and the ∆5 form in the brain and cell lines. Therefore, both the long/short and the ∆5 alternative splicing were tissue specific.

CPEB1 localization in cytoplasmic foci In order to study the intracellular localization of the CPEB1 protein, monoclonal antibodies were raised against the human CPEB1-∆5 protein fused to GST and produced in E. coli. Their specificity was demonstrated by western blotting using extracts of HeLa cells transfected with CPEB1 expression vectors (Fig. 2A). As an example, the 2B7 antibody identified both the Journal of Cell Science human long and mouse short CPEB1-∆5 proteins at sizes consistent with their predicted molecular weight of 62 and 53 kDa, respectively. This antibody was then used to determine the intracellular localization of endogenous CPEB1 protein in HeLa cells by immunofluorescence (Fig. 2B). The staining was weak and concentrated in a few discrete cytoplasmic foci. Fig. 3. Localization of GFP-tagged CPEB1 in HeLa cells. The specificity of these immunofluorescence data was (A,B) Localization of CPEB1 after transient transfection. HeLa cells were transiently transfected with an expression vector for GFP-tagged confirmed using GFP-tagged CPEB1 protein. An open reading ∆ ∆ human CPEB1- 5-long. After 24 hours, cells were fixed, stained with frame encoding human CPEB1- 5-long fused to GFP at the C- DAPI and observed by fluorescence microscopy. CPEB fluorescence terminus, was introduced into a plasmid under the control of a and DAPI staining are on the left and right, respectively. Cells CMV promoter and transiently transfected into HeLa cells. In harbouring small cytoplasmic foci (arrows) and larger granules are about two thirds of the cells, CPEB1-∆5 exhibited a illustrated in A and B, respectively. (C,D) Localization of CPEB1 after cytoplasmic diffuse localization with a few discrete enriched stable transfection. HeLa/CPEB1-lg cells were induced for 16 hours foci (Fig. 3A), as observed with the anti-CPEB1 antibody in with doxycycline, fixed and stained with DAPI (C) or observed live untransfected HeLa cells. In the remaining cells, the protein (D). The arrows indicate small cytoplasmic foci. Bar, 8 µm. was concentrated in larger granules (Fig. 3B). The same open reading frame was also inserted into a However, the larger granules previously observed in transient plasmid under the control of a cytomegalovirus/tet promoter transfection were absent. and transfected into the HeLa rtTA HR5 cell line to allow for inducible expression. Stable transfectants were selected and a clone with a high level of induction by doxycycline, called Induction of stress granules by CPEB HeLa/CPEB1-lg, was chosen for further study. In every We first analysed the large CPEB1 granules observed following induced cell, the intracellular localization of the GFP-tagged transient transfection. Their morphology was similar to the CPEB1-∆5 protein consisted of a diffuse cytoplasmic staining previously described cytoplasmic stress granules (Kedersha et as well as a few discrete cytoplasmic foci (Fig. 3C). This al., 2002). These granules can be induced by stress such as localization in foci was also observed in living cells (Fig. 3D). arsenite, UV or heat shock, as well as by expression of a dcp1 bodies and stress granules 985

Fig. 4. Large CPEB1 granules are stress granules. HeLa cells were transiently transfected with an expression vector for GFP-tagged human CPEB1- ∆5-long. After 24 hours, cells were directly fixed (A) or stressed with arsenite for 30 minutes and fixed (B), then stained with anti-eIF3 antibodies (red) and observed by fluorescence microscopy.

phosphomimetic mutant of the translation initiation factor eIF2-α. They contain many translation initiation factors, in particular eIF3, mRNAs and some mRNA binding proteins. HeLa cells were transiently transfected with the GFP-tagged CPEB1-∆5-long expression vector and analysed with an anti- eIF3 antibody. The localization of eIF3 was diffuse in untransfected cells as well as in those where the CPEB1 protein was diffuse. However, when cells harboured larger CPEB1 granules, eIF3 was enriched in these granules (Fig. 4A). In addition, transfected cells were stressed with arsenite in order to induce stress granule-like structures,as does a phosphomimetic mutant of granules. As previously described (Kedersha et al., 2002), all eIF2-α. cells harboured eIF3-containing granules. Endogenous CPEB1 became undetectable using our antibodies (see later). However, the transfected CPEB1 protein was systematically enriched in Cytoplasmic CPEB1 small foci are dependent on these granules (Fig. 4B). The same result was obtained when translation the stably transfected HeLa/CPEB1-lg cells were induced and We then turned to the smaller CPEB1 cytoplasmic foci. As stressed with arsenite (data not shown). CPEB1 protein regulates polyadenylation and translation, In conclusion, CPEB1 protein was recruited into stress we investigated whether the presence of these foci was granules during stress, and its transient overexpression in dependent upon active translation. HeLa/CPEB1-lg cells were unstressed cells was able to induce the formation of stress treated with two translation inhibitors, cycloheximide and puromycin. After 1 hour neither the cycloheximide nor the puromycin Journal of Cell Science had notably decreased the abundance of the diffuse CPEB1 protein. However, there was a striking difference in the small cytoplasmic foci (Fig. 5A). They disappeared in the presence of cycloheximide, whereas puromycin enhanced their intensity and increased their number. The number of cells with more than five foci rose from 3% to 37% and the number of cells without any foci decreased from 40% to 10% (Fig. 5B). These results indicated that CPEB1 foci are dynamic structures. The mechanism of action of the two translation inhibitors is different: puromycin leads to the disruption of polysomes and to the release of both the nascent peptides and the mRNAs, whereas cycloheximide traps arrested mRNA on polysomes. As Fig. 5. Small CPEB1 foci depend on translation. (A) HeLa/CPEB1-lg cells were induced for 16 the number of CPEB1 foci increased hours with doxycycline, treated with cycloheximide (CHX) for 40 minutes or puromycin for 1 after puromycin treatment, they hour, then fixed and observed by fluorescence microscopy. (B) The number of CPEB1 foci per cell could not correspond to structures was counted in 150 control, cycloheximide- and puromycin-treated cells. involved in active translation. In 986 Journal of Cell Science 118 (5) mammals, the decapping enzyme dcp1 (Ingelfinger et al., 2002; van Dijk et al., 2002) and the mRNA binding protein GW182 (Eystathioy et al., 2002) specifically accumulate in dcp1 bodies. To determine whether the CPEB1 foci were related to dcp1 bodies, we induced HeLa/CPEB1- lg cells with doxycycline and performed immunofluorescence experiments with either anti- dcp1 or anti-GW182 antibodies. The CPEB1 cytoplasmic foci fully colocalized with both dcp1 (Fig. 6A) and GW182 foci (Fig. 6B). The same results were obtained when the unfused human CPEB1-∆5-long expression vector was transiently transfected into HeLa cells and stained with the anti-CPEB1 antibody (Fig. 6C), ensuring that the colocalization with dcp1 and GW182 was not due to the GFP moiety of the recombinant CPEB1 expressed in HeLa/CPEB1-lg cells. Finally, the endogenous CPEB1 protein detected with the anti- CPEB1 antibody also colocalized with the dcp1 protein in untransfected HeLa cells (Fig. 6D). In conclusion, the small CPEB1 foci corresponded to the previously described dcp1 bodies.

Functional RRMs are required for CPEB1 targeting to both stress granules and dcp1 bodies The existence of several mammalian CPEB1 isoforms, CPEB1-long and -short, ∆5 or not, raised the possibility of multiple functions for CPEB1, possibly related to distinct localizations. To address this question, the four isoforms were introduced as GFP fusions into expression vectors and transiently transfected in HeLa cells. All of Journal of Cell Science them had a similar cytoplasmic localization, with presence of the protein in either stress granules (Fig. 7A) or dcp1 bodies (Fig. 7B), as confirmed with anti-eIF3 and anti-dcp1 antibodies, respectively. Therefore, neither the first 68 amino acids nor the GNMPK alternative motif were involved in the stress granule assembly or in the Fig. 6. Small CPEB1 foci are dcp1/GW bodies. HeLa/CPEB1-lg cells were induced localization in dcp1 bodies. for 16 hours with doxycycline, fixed and stained with anti-dcp1 (A) or anti-GW182 As a first approach to determine the functional (B) antibodies (red). HeLa cells transiently transfected with an expression vector ∆ domain involved in these localizations, two large for untagged human CPEB1- 5-long (C) and untransfected HeLa cells (D) were deletions were created in the GFP-tagged CPEB1- fixed and stained with anti-dcp1 (red) and anti-CPEB1 (green) antibodies. DAPI staining is blue. long expression vector, one removing the amino acids 1 to 237 at the N-terminus, and the other amino acids 330 to 566 at the C-terminus of the contrast, the effect of cycloheximide indicated that these foci protein. The former deletion encompassed the Aurora A depended on the ability of mRNAs to leave polysomes at the phosphorylation site and the PEST region, whereas the latter issue of translation. removed the RRMs and Zn finger so that the protein could not bind RNA (Hake et al., 1998). The N-terminal truncated protein was identical to the full-length protein in its capacity to Cytoplasmic CPEB1 foci are dcp1 bodies assemble stress granules. It was also recruited into dcp1 bodies, These characteristics were reminiscent of the properties of dcp1 although with less efficiency than the full-length protein (data bodies in mammals (Cougot et al., 2004) and P-bodies in yeast not shown). In contrast, neither stress granules nor smaller (Sheth and Parker, 2003), two cytoplasmic structures involved CPEB1 foci were observed with the C-terminal truncated in mRNA degradation. They have been shown to disappear mutant, despite the presence of proper dcp1 foci within the cells following cycloheximide treatment and to be enhanced (Fig. 7C). Moreover, this mutant was not recruited into stress following the disruption of the mRNA degradation pathway. In granules following arsenite treatment (Fig. 7D). dcp1 bodies and stress granules 987 Journal of Cell Science Fig. 7. Stress granule assembly and dcp1 bodies depend on the functionality of the mRNA binding domain. (A) The four CPEB1 isoforms can assemble stress granules. HeLa cells were transiently transfected with expression vectors for the indicated GFP-tagged CPEB1 isoforms. After 24 hours, cells were fixed, stained with anti-eIF3 antibodies (red) and observed by confocal microscopy. The figure illustrates only cells harbouring stress granules. (B) All CPEB1 isoforms can colocalize with dcp1 bodies. Cells obtained as in A were stained with anti-dcp1 (red) and observed by confocal microscopy. The figure illustrates only cells harbouring small CPEB1 foci. (C) A C-terminal truncation of CPEB1 abrogates the colocalization with dcp1 bodies. HeLa cells were transiently transfected with an expression vector for an RRM- and Zn finger- deleted CPEB1-long. After 24 hours, cells were fixed and stained with anti-dcp1 antibodies (red). (D) A C-terminal truncation of CPEB1 abrogates the localization in stress granules. HeLa cells transfected as in C were stressed with arsenite for 30 minutes, fixed and stained with anti-eIF3 antibodies (red). (E) H545A and F314A point mutations abrogate the localization of CPEB1 in dcp1 bodies. HeLa cells were transfected with an expression vector for GFP-tagged CPEB1-long-H545A or -F314A and stressed as in D. Cells were stained with anti-dcp1 antibodies (red).

To investigate in more detail the role of Aurora A truncated mutant, harbouring neither dcp1 bodies nor stress phosphorylation and mRNA binding, we introduced a set of granules (Fig. 7E). specific point mutations. The Thr172 targeted by Aurora A was Therefore both stress granule assembly and localization in mutated either to Asp, known as a phosphomimetic mutation, dcp1 bodies required a functional mRNA binding domain, or to Ala, a non-phosphorylatable amino acid. None of these whereas the Aurora A phosphorylation site and the PEST mutations affected the localization of CPEB1 in the cells (data domain were dispensable. not shown), confirming that Aurora A phosphorylation is not involved here. For inhibition of mRNA binding, we created two independent mutations. One was a His545 to Ala mutation Stress granules recruit dcp1 bodies within the zinc finger, previously shown to inactivate the The fact that the CPEB1 protein could both accumulate in dcp1 mRNA binding domain of Xenopus CPEB (Hake et al., 1998). bodies and assemble stress granules, two structures involved in The other is a Phe314 to Ala mutation in the rnp2 motif of mRNA storage and/or degradation, led us to investigate the RRM1, known to abolish mRNA binding in RRM-containing relationship between the two. HeLa cells were transiently proteins (Jessen et al., 1991). Both behaved like the C-terminal transfected with GFP-tagged CPEB1-∆5-short to induce stress 988 Journal of Cell Science 118 (5) granules and analysed for dcp1 localization after 20 hours. Several patterns were observed in the stress granule-containing cells. In 70% of cells, the punctuated dcp1 bodies had disappeared and dcp1 protein was present in the stress granules (Fig. 8A). In 10% of cells, punctuated dcp1 bodies coexisted with and were distinct from stress granules (Fig. 8B). In the remaining 20%, dcp1 bodies were distinct but closely associated with stress granules. There were often two to three bodies contacting one stress granule (Fig. 8C). The presence of GW182 in these cells was also analysed. Similar to dcp1, GW182 bodies had disappeared and the protein accumulated in stress granules (Fig. 8D). Therefore, stress granules induced by CPEB1 recruited two components of dcp1 bodies. Although the two proteins may join the stress granules separately, the image of dcp1 bodies surrounding stress granules suggested rather that dcp1 bodies were recruited as a whole by stress granules. It was recently reported that dcp1 is absent from stress granules induced with arsenite in HEK293 cells (Cougot et al., 2004). We analysed dcp1 localization in our HeLa/CPEB1-lg cells treated with arsenite for 30 minutes. As in HEK293 cells, dcp1 was not recruited into newly assembled stress granules (data not shown). After recovery for 1 hour 45 minutes in the absence of arsenite, dcp1 was still not recruited and after 2 hours 30 minutes, stress granules had disappeared. It was interesting that the number of dcp1 bodies strongly increased in response to arsenite (Fig. 8E), which is consistent with its inhibitory effect on translation. It also suggests that, in this case, many repressed mRNAs are still targeted to dcp1 bodies rather than stress granules. The fact that dcp1 joins CPEB1-induced but not arsenite- induced stress granules may be due to a difference in the timing of the observation. We therefore analysed dcp1 localization during shorter time of transfection. GFP-tagged CPEB1 expression became visible after 8 hours of transfection and cells Journal of Cell Science were observed up to 24 hours. Among transfected cells, the number of stress granule-containing cells increased with time

Fig. 8. Stress granules induced by CPEB1, but not by arsenite, recruit components of dcp1 bodies. (A-D) CPEB1-induced stress granules recruit dcp1 and GW182. HeLa cells were transiently transfected with an expression vector for GFP-tagged CPEB1-∆5- short. After 20 hours, cells were fixed and stained with anti-dcp1 (A-C) or anti-GW182 (D) antibodies (red). Cells were observed by confocal microscopy. As several patterns were observed, three cells are shown in A, B and C, which are representative of 70%, 10% and 20% of the stress granule-containing cells, respectively. In C, two stress granules surrounded with several dcp1 bodies have been enlarged for clearer visualization. (E) Arsenite increases the number of dcp1 bodies. HeLa/CPEB1-lg cells were induced for 16 hours with doxycycline, stressed with arsenite (as) for 30 minutes, or stressed and cultured further in the absence of arsenite for 1 hour (as +1 hour). After fixation, cells were stained with anti-dcp1 antibodies and observed by fluorescence microscopy. The graph presents the number of dcp1 bodies per cell, counted in 180 cells. (F) Recruitment of dcp1 in CPEB1-induced stress granules increases with time. HeLa cells transfected as in A were fixed at various time after transfection and stained with anti-dcp1 antibodies. CPEB1- expressing cells were counted for the presence of stress granules (SG) and the presence of dcp1 in these stress granules. Stress granule-containing cells are plotted as a percentage of CPEB1- expressing cells, whereas cells with dcp1 in stress granules are plotted as a percentage of stress granule-containing cells. dcp1 bodies and stress granules 989

Fig. 9. Human p54 colocalizes with CPEB1 in both dcp1 and stress granules. (A) p54 is found in dcp1 bodies. HeLa cells were transiently transfected with an expression vector for RFP-tagged p54. After 20 hours, cells were fixed, stained with anti-dcp1 antibodies (green), and observed by confocal microscopy. (B) p54 colocalizes with CPEB1. HeLa cells were cotransfected with RFP-tagged p54 and GFP-tagged CPEB1-long. After 20 hours, cells were fixed and observed by confocal microscopy. Cells without (upper panel) and with stress granules (lower panel) are presented. (C,D) p54 is recruited with endogenous CPEB1 in stress granules induced with arsenite. HeLa cells were transiently transfected with an expression vector for RFP-tagged p54. After 20 hours, cells were stressed with arsenite for 30 minutes, incubated in the absence of arsenite for 1 hour and fixed. After staining with anti- eIF3 (C) or anti-CPEB1 antibodies (D) (green), cells were observed by confocal microscopy. In C, two p54 foci contacting stress granules have been enlarged for better visualization.

(Fig. 8F). In addition, among the latter, the number of cells where dcp1 had joined stress granules also increased with time (Fig. 8F). This indicated that stress granules take time to assemble and need further time to recruit dcp1 bodies.

The rck/p54 protein is colocalized with CPEB1 in both stress granules and dcp1 bodies The DEAD box helicase rck/p54 is known as a CPEB partner in Xenopus and clam (Minshall and Standart, 2004; Minshall et al., 2001). It is also a component of P-bodies (Sheth and Parker, 2003) and dcp1 bodies (Cougot et al., 2004). We therefore Journal of Cell Science analysed its localization following CPEB1 transfection and arsenite treatment. An expression vector for RFP-tagged human p54 was constructed and transfected into HeLa cells. As expected, the fusion protein concentrated in dcp1 bodies, as confirmed by dcp1 colocalization (Fig. 9A), and did not trigger stress granule assembly. When RFP-tagged p54 was cotransfected with GFP-tagged CPEB1-long the protein systematically colocalized with CPEB1, either in dcp1 bodies (Fig. 9B, upper panel) or in CPEB1-induced stress granules (Fig. 9B, lower panel). Finally, p54 was transfected alone for 20 hours and stress granules were induced with arsenite. In half of the cells, p54 was found in stress granules, as confirmed by anti- eIF3 antibodies (Fig. 9C, middle panel). In the remaining cells, it was present in punctuated dcp1 bodies. These bodies were distinct from but often associated with stress granules (Fig. 9C, lower panel). We analysed endogenous CPEB1 in these cells using our anti-CPEB1 antibody. In contrast to untransfected cells, the endogenous protein was now detectable in the stress granules (Fig. 9D). Therefore, p54 is colocalized with CPEB1 in dcp1 bodies, as well as in stress granules, whether induced by CPEB1 or arsenite. For comparison, 990 Journal of Cell Science 118 (5)

Stress mutant eIF2α Fig. 10. Proposed model for the link between stress CPEB-GFP granules and mRNA degradation bodies. Translation FMRP block in response to stress or translational G3BP stress repression by factors such as FMRP, mutated eIF2-α granule or GFP-CPEB, leads to the storage of mRNAs in stress granules. These granules can either revert if mRNA the stress disappears or progressively recruit dcp1 storage stress stress bodies to degrade mRNAs. In contrast, translation granule granule inhibitors enabling the release of mRNAs, like puromycin, lead directly to the default mRNA translation mRNA degradation pathway which takes place within the degradation dcp1 bodies. However, translation inhibitors that dcp1 body trap arrested mRNAs on polysomes, such as CHX mRNA cycloheximide (CHX), prevent them from joining Puro mRNA degradation the dcp1 bodies.

dcp1 is recruited in only 70% of the CPEB1-induced stress transfected HeLa cells correspond to so-called stress granules. granules and is absent in arsenite-induced stress granules. These granules are formed in response to stress such as arsenite, UV and heat shock, but can also be induced by the expression of a phosphomimetic mutant of eIF2-α (Kedersha Discussion et al., 2002). Their assembly results from the accumulation of Two CPEB1 isoforms have been described in mammals, abortive translation preinitiation complexes, which are CPEB1-long and -short, resulting from the alternative splicing deficient in eIF2/eIF5 (Anderson and Kedersha, 2002). As for of the first coding exon. Here we show the existence of a the mutated eIF2-α, the overexpression of GFP-tagged CPEB1 second alternative splicing, which leads to a deletion of five induced the assembly of granules that contain the translation amino acids within the RRM1. This new isoform, named initiation factor eIF3, a marker of stress granules (Kedersha et CPEB1-∆5, was also found among zebrafish ESTs. Both the al., 2002). Conversely, stress granules induced by arsenite (Fig. long/short and the ∆5 alternative splicing were differentially 4), UV or heat shock (data not shown) recruit CPEB1 protein. regulated in mouse brain and ovary. The long form was In Xenopus oocytes, CPEB1 represses translation initiation of common to both tissues, whereas the short form was its target mRNAs by preventing the binding of eIF4G to eIF4E. exclusively found in the brain. In addition, the non-∆5 and the In mammalian cells, an excess of CPEB1 could repress ∆5 forms were predominant in the ovary and in the brain, translation, leading to the assembly of defective initiation respectively. This was in agreement with the single CPEB complexes, which would accumulate in stress granules, isoform cloned in Xenopus oocytes, which is long and non-∆5. enabling the storage or the degradation of blocked mRNAs. Journal of Cell Science In two human cell lines, HeLa and RPE-1, only the long form Stress granule inducers reported to date are: a phosphomimetic was detected and predominantly the ∆5 form, as in the mouse mutant of eIF2-α, a factor which belongs to the translation brain. initiation complex (Kedersha et al., 2002), and two RNA The five amino acids that are deleted in the CPEB1-∆5 binding proteins, the Fragile X Mental Retardation Protein isoform are located within the RRM1, at the C-terminal (FMRP), which represses translation (Mazroui et al., 2002) and extremity of the β2-β3 loop. It is noteworthy that, for U1A, sx1 the Ras-GAP SH3 domain-binding protein (G3BP), which has and nucleolin, amino acids at this position have been shown to an endoribonuclease activity (Tourriere et al., 2003). Thus interact directly with RNA of the RRM (Allain et al., 2000; CPEB1 is the second example of a translational regulator that Handa et al., 1999). A deletion in this region could therefore can induce stress granules and the efficiency of this induction affect the RNA binding properties of CPEB1, at the affinity or could indicate that CPEB1 has many more target mRNAs than specificity level. Its differential expression in ovary and brain anticipated. could correspond to the existence of different targets in these The small CPEB1 foci observed in stably and transiently tissues, for instance mitosis-related and signal transduction- CPEB1-transfected cells correspond to previously reported related mRNAs, as postulated for oocyte and neurons, dcp1 bodies. These cytoplasmic bodies have been described in respectively (Wells et al., 2000). both yeast and mammalian cells. They contain factors The intracellular localization of the CPEB1 protein was participating in the 5′-3′ mRNA decay pathway, including the investigated in HeLa cells using a monoclonal antibody decapping enzymes dcp1 and dcp2, the exonuclease Xrn1, the developed in the laboratory. The protein was detected in small Sm-like proteins LSm1 to 7, the deadenylase hCcr4 and the cytoplasmic foci. To distinguish between the four CPEB1 helicase dhh1/rck/p54 (Cougot et al., 2004; Ingelfinger et al., isoforms, expression vectors for GFP-tagged CPEB1 and 2002; Sheth and Parker, 2003; van Dijk et al., 2002). They also CPEB1-∆5, -long and -short, were transiently transfected in contain an RNA binding protein of unknown function, GW182, HeLa cells. The four isoforms had a similar diffuse initially identified as a human autoantigen (Eystathioy et cytoplasmic localization, with either small, dispersed foci or al., 2002; Eystathioy et al., 2003). CPEB1 foci were much larger granules that were irregular in size. In stably indistinguishable from dcp1 or GW182 foci. transfected HeLa cells, only the first pattern was observed. In yeast, these foci have been shown to be mRNA The large CPEB1 granules observed in transiently degradation sites rather than storage sites (Sheth and Parker, dcp1 bodies and stress granules 991 2003). In particular, mutations that block decapping or 5′-3′ distinct stress granules and dcp1 bodies up to the time at which mRNA degradation increase the number and size of these foci, stress granules disappear, i.e. 2.5 hours. In contrast, when stress whereas translation inhibition by cycloheximide, which traps granules were induced by CPEB1 expression, the typical dcp1 mRNAs on polysomes, causes their disappearance. Here, we bodies had disappeared in most cells and the dcp1 and GW182 provide similar functional data in mammalian cells. On the one proteins previously present in these bodies filled the stress hand, mRNA release from polysomes by puromycin treatment granules. This difference between arsenite-induced and led to a strong increase in the number of foci. On the other CPEB1-induced stress granules may result from a difference hand, blocking mRNAs on polysomes by cycloheximide in the timing of the observation. The former were observed treatment led to the disappearance of CPEB1 foci, as in yeast. shortly after assembly, whereas the latter were observed after These observations are consistent with a recent publication 20 hours, allowing more time to recruit further factors. reporting that reducing 5′-3′ mRNA degradation by Xrn-1 Supporting that hypothesis, we found that the recruitment of silencing leads to an increased number of dcp2 bodies, whereas dcp1 into stress granules strongly increases between 8 hours cycloheximide treatment diminishes their number (Cougot et and 20 hours after transfection. Interestingly, intermediate al., 2004). Together, these data demonstrate that these foci are situations were observed, where the dcp1 bodies had not yet dynamic and correspond to active sites of mRNA degradation. disappeared, but were adjacent to the stress granules. There The CPEB1 protein can be seen as both a translational was often more than one dcp1 body per stress granule, as if activator when phosphorylated and recruiting the stress granules would progressively recruit and fuse with dcp1 polyadenylation complex, and a translational inhibitor when bodies. When p54 was overexpressed in the cells, it was dephosphorylated and bound to a protein such as maskin, recruited in CPEB1-induced stress granules, like dcp1 and which interferes with eIF4E function in translation initiation GW182. Moreover, it could also be recruited in arsenite- (Stebbins-Boaz et al., 1999). We postulate that the presence of induced stress granules, along with endogenous CPEB1. CPEB1 in dcp1 bodies is probably related to its function Again, when p54 was still in dcp1 bodies, these bodies were as a translational inhibitor. Accordingly, the Aurora A most often adjacent to stress granules. Therefore, the phosphorylation site, which controls CPEB activity as a composition of arsenite-induced stress granules is highly translational activator, is not required for this localization. The dependent on the cellular content. A high expression of p54 capacity of the four CPEB1 isoforms to associate with these seems to accelerate the recruitment of dcp1 bodies. This bodies was indistinguishable and therefore did not suggest any communication between stress granules and dcp1 bodies is the difference in their respective inhibitor activity. The presence of first direct evidence for the postulated role of stress granules, CPEB1 in the bodies did require a functional mRNA binding which is to sort mRNAs for storage or degradation. domain, indicating that the protein is bound to its target mRNAs within them. Most proteins documented so far in dcp1 We thank Michèle Huesca for raising monoclonal antibodies, bodies are proteins directly involved in mRNA degradation. An Marie-Annick Harper for technical help and Gérard Pierron for RNA binding protein involved in translational repression such reading of the manuscript. A.W. was supported by a Marie Curie as CPEB1 could trigger the recruitment of its target mRNAs Fellowship (5th PCRD, contract QLGA-1999-50406) and a FEBS Collaborative Experimental Scholarship for Central and Eastern to dcp1 bodies and, as a consequence, be dragged into them.

Journal of Cell Science Europe. This work was supported by the Centre National de la Along these lines, an MS2-GFP tag, carried along by an mRNA Recherche Scientifique and the Marie Curie Fellowship. containing MS2 binding sites, was visualized in P-bodies, provided that the mRNA was enriched in that location by inhibition of either mRNA translation or mRNA degradation References (Sheth and Parker, 2003). However, it may not be so simple, Aakalu, G., Smith, W. B., Nguyen, N., Jiang, C. and Schuman, E. M. (2001). Dynamic visualization of local protein synthesis in hippocampal as the yeast Puf3p protein, albeit a regulator of mRNA neurons. Neuron 30, 489-502. degradation, is not enriched in dcp1 bodies (Sheth and Parker, Allain, F. H., Gilbert, D. E., Bouvet, P. and Feigon, J. (2000). Solution 2003). Alternatively, the accumulation of CPEB1 in these structure of the two N-terminal RNA-binding domains of nucleolin and bodies could reveal an active role for the protein in mRNA NMR study of the interaction with its RNA target. J. Mol. Biol. 303, 227- degradation. In this respect, it is of interest that mammalian 241. Anderson, P. and Kedersha, N. (2002). Stressful initiations. J. Cell Sci. 115, rck/p54 accumulates in dcp1 bodies. Its homologues in the 3227-3234. Xenopus and clam (Xp54 and p47, respectively) are known to Audibert, A., Weil, D. and Dautry, F. (2002). In vivo kinetics of mRNA be CPEB partners and translational repressors during early splicing and transport in mammalian cells. Mol. Cell Biol. 22, 6706-6718. development (Minshall and Standart, 2004; Minshall et al., Blochlinger, K. and Diggelmann, H. (1984). Hygromycin B phosphotransferase as a selectable marker for DNA transfer experiments 2001). Its homologue in yeast, Dhh-1, binds to dcp1 and is with higher eucaryotic cells. Mol. Cell Biol. 4, 2929-2931. required for efficient decapping (Coller et al., 2001; Fischer Coller, J. M., Tucker, M., Sheth, U., Valencia-Sanchez, M. A. and Parker, and Weis, 2002). This raises the possibility that CPEB1 is R. (2001). The DEAD box helicase, Dhh1p, functions in mRNA decapping bound to rck/p54 within the dcp1 body and plays an active role and interacts with both the decapping and deadenylase complexes. RNA 7, in decapping. 1717-1727. Cougot, N., Babajko, S. and Seraphin, B. (2004). Cytoplasmic foci are sites Finally, our results illustrate a link between stress granules of mRNA decay in human cells. J. Cell Biol. 165, 31-40. and dcp1 bodies, as schematically proposed in Fig. 10. It has Dirks, W., Schaper, F., Kirchhoff, S., Morelle, C. and Hauser, H. (1994). been recently reported that stress granules newly assembled in A multifunctional vector family for gene expression in mammalian cells. arsenite-treated cells and dcp1 bodies do not overlap, although Gene 149, 387-388. Eystathioy, T., Chan, E. K., Tenenbaum, S. A., Keene, J. D., Griffith, K. they are occasionally close together (Cougot et al., 2004). and Fritzler, M. J. (2002). A phosphorylated cytoplasmic autoantigen, These authors concluded that the two structures were distinct. GW182, associates with a unique population of human mRNAs within novel Similarly, when treating cells with arsenite, we observed cytoplasmic speckles. Mol. Biol. Cell 13, 1338-1351. 992 Journal of Cell Science 118 (5)

Eystathioy, T., Jakymiw, A., Chan, E. K., Seraphin, B., Cougot, N. and Mendez, R., Murthy, K. G., Ryan, K., Manley, J. L. and Richter, J. D. Fritzler, M. J. (2003). The GW182 protein colocalizes with mRNA (2000). Phosphorylation of CPEB by Eg2 mediates the recruitment of CPSF degradation associated proteins hDcp1 and hLSm4 in cytoplasmic GW into an active cytoplasmic polyadenylation complex. Mol. Cell 6, 1253- bodies. RNA 9, 1171-1173. 1259. Fischer, N. and Weis, K. (2002). The DEAD box protein Dhh1 stimulates the Minshall, N. and Standart, N. (2004). The active form of Xp54 RNA helicase decapping enzyme Dcp1. EMBO J. 21, 2788-2797. in translational repression is an RNA-mediated oligomer. Nucleic Acids Res. Gebauer, F. and Richter, J. D. (1996). Mouse cytoplasmic polyadenylylation 32, 1325-1334. element binding protein: an evolutionarily conserved protein that interacts Minshall, N., Thom, G. and Standart, N. (2001). A conserved role of a with the cytoplasmic polyadenylylation elements of c-mos mRNA. Proc. DEAD box helicase in mRNA masking. RNA 7, 1728-1742. Natl. Acad. Sci. USA 93, 14602-14607. Sambrook, J., Fritsch, E. F. and Maniatis, T. (1989). Molecular Cloning. Groisman, I., Huang, Y. S., Mendez, R., Cao, Q., Theurkauf, W. and Cold Spring Harbour, Cold Spring Harbor Laboratory Press. Richter, J. D. (2000). CPEB, maskin, and cyclin B1 mRNA at the mitotic Sheth, U. and Parker, R. (2003). Decapping and decay of messenger RNA apparatus: implications for local translational control of cell division. Cell occur in cytoplasmic processing bodies. Science 300, 805-808. 103, 435-447. Stebbins-Boaz, B., Cao, Q., de Moor, C. H., Mendez, R. and Richter, J. D. Hake, L. E., Mendez, R. and Richter, J. D. (1998). Specificity of RNA (1999). Maskin is a CPEB-associated factor that transiently interacts with binding by CPEB: requirement for RNA recognition motifs and a novel zinc elF-4E. Mol. Cell 4, 1017-1027. finger. Mol. Cell Biol. 18, 685-693. Theis, M., Si, K. and Kandel, E. R. (2003). Two previously undescribed Handa, N., Nureki, O., Kurimoto, K., Kim, I., Sakamoto, H., Shimura, Y., members of the mouse CPEB family of genes and their inducible expression Muto, Y. and Yokoyama, S. (1999). Structural basis for recognition of the in the principal cell layers of the hippocampus. Proc. Natl. Acad. Sci. USA tra mRNA precursor by the Sex-lethal protein. Nature 398, 579-585. 100, 9602-9607. Huang, Y. S., Jung, M. Y., Sarkissian, M. and Richter, J. D. (2002). N- Tourriere, H., Chebli, K., Zekri, L., Courselaud, B., Blanchard, J. M., methyl-D-aspartate receptor signaling results in Aurora kinase-catalyzed Bertrand, E. and Tazi, J. (2003). The RasGAP-associated CPEB phosphorylation and alpha CaMKII mRNA polyadenylation at endoribonuclease G3BP assembles stress granules. J. Cell Biol. 160, 823- synapses. EMBO J. 21, 2139-2148. 831. Ingelfinger, D., Arndt-Jovin, D. J., Luhrmann, R. and Achsel, T. (2002). van Dijk, E., Cougot, N., Meyer, S., Babajko, S., Wahle, E. and Seraphin, The human LSm1-7 proteins colocalize with the mRNA-degrading enzymes B. (2002). Human Dcp2: a catalytically active mRNA decapping enzyme Dcp1/2 and Xrnl in distinct cytoplasmic foci. RNA 8, 1489-1501. located in specific cytoplasmic structures. EMBO J. 21, 6915-6924. Jessen, T. H., Oubridge, C., Teo, C. H., Pritchard, C. and Nagai, K. (1991). Weil, D., Boutain, S., Audibert, A. and Dautry, F. (2000). Mature Identification of molecular contacts between the U1 A small nuclear mRNAs accumulated in the nucleus are neither the molecules in transit to ribonucleoprotein and U1 RNA. EMBO J. 10, 3447-3456. the cytoplasm nor constitute a stockpile for gene expression. RNA 6, 962- Kedersha, N., Chen, S., Gilks, N., Li, W., Miller, I. J., Stahl, J. and 975. Anderson, P. (2002). Evidence that ternary complex (eIF2-GTP- Welk, J. F., Charlesworth, A., Smith, G. D. and MacNicol, A. M. (2001). tRNA(i)(Met))-deficient preinitiation complexes are core constituents of Identification and characterization of the gene encoding human cytoplasmic mammalian stress granules. Mol. Biol. Cell 13, 195-210. polyadenylation element binding protein. Gene 263, 113-120. Kurihara, Y., Tokuriki, M., Myojin, R., Hori, T., Kuroiwa, A., Matsuda, Wells, D. G., Richter, J. D. and Fallon, J. R. (2000). Molecular mechanisms Y., Sakurai, T., Kimura, M., Hecht, N. B. and Uesugi, S. (2003). CPEB2, for activity-regulated protein synthesis in the synapto-dendritic a novel putative translational regulator in mouse haploid germ cells. Biol. compartment. Curr. Opin. Neurobiol. 10, 132-137. Reprod. 69, 261-268. Wu, L., Wells, D., Tay, J., Mendis, D., Abbott, M. A., Barnitt, A., Mazroui, R., Huot, M. E., Tremblay, S., Filion, C., Labelle, Y. and Quinlan, E., Heynen, A., Fallon, J. R. and Richter, J. D. (1998). CPEB- Khandjian, E. W. (2002). Trapping of messenger RNA by Fragile X Mental mediated cytoplasmic polyadenylation and the regulation of experience- Retardation protein into cytoplasmic granules induces translation repression. dependent translation of alpha-CaMKII mRNA at synapses. Neuron 21, Hum. Mol. Genet. 11, 3007-3017. 1129-1139. Journal of Cell Science