Mammalian Smaug Is a Translational Repressor That Forms Cytoplasmic Foci Similar to Stress Granules*

Cytoplasmic events depending on RNA-binding proteins contribute to the fine-tuning of gene expression. Sterile α motif-containing RNA-binding proteins constitute a novel family of post-transcriptional regulators that recognize a specific RNA sequence motif known as Smaug recognition element (SRE). The Drosophila member of this family, dSmaug, triggers the translational repression and deadenylation of maternal mRNAs by independent mechanisms, and the yeast homologue Vts1 stimulates degradation of SRE-containing messengers. Two homologous genes are present in the mammalian genome. Here we showed that hSmaug 1, encoded in human chromosome 14, represses the translation of reporter transcripts carrying SRE motifs. When expressed in fibroblasts, hSmaug 1 forms cytoplasmic granules that contain polyadenylated mRNA and the RNA-binding proteins Staufen, TIAR, TIA-1, and HuR. Smaug 1 foci are distinct from degradation foci. The murine protein mSmaug 1 is expressed in the central nervous system and is abundant in post-synaptic densities, a subcellular region where translation is tightly regulated by synaptic stimulation. Biochemical analysis indicated that mSmaug 1 is present in synaptoneurosomal 20 S particles. These results suggest a role for mammalian Smaug 1 in RNA granule formation and translation regulation in neurons.

Messenger RNA localization, translation activation, silencing, and controlled degradation contribute to the fine-tuning of gene expression in time and space. All these processes depend on several families of RNA-binding proteins that are of comparable importance to transcription factors in regulating gene expression (1). Sterile ␣ motif (SAM) 3 containing RNA binding domains define a novel family of RNA-binding proteins that function as post-transcriptional regulators (2). They bind to an RNA sequence motif known as SRE (Smaug recognition element), the Drosophila protein Smaug being the first member that was identi-fied (2)(3)(4)(5). Drosophila Smaug is involved in translational repression of the maternal mRNA encoding nanos, a posterior determinant, and thus plays a role in defining embryo polarity. Smaug recruits Cup, an eIF4Ebinding protein that prevents the association of eIF4E with eIF4G, thus blocking initiation of the translation of SRE-containing messengers (6). In addition, it has been reported recently that Drosophila Smaug mediates degradation of maternal Hsp83 mRNAs by an independent mechanism that involves the CCR4 deadenylase and does not require Cup nor SRE motifs (7). The yeast homologue Vts1 stimulates degradation of SRE-containing messengers by a similar mechanism (2).
Two Smaug homologous genes of unknown function are present in the mammalian genome (2,3). Here we show that Smaug 1, encoded in human chromosome 14, represses translation of SRE-containing messengers in fibroblast cell lines. Both hSmaug 1 and Drosophila Smaug form cytoplasmic granules when expressed in fibroblasts and colocalize when cotransfected. Furthermore, hSmaug 1 foci contain polyadenylated mRNAs, and their size and number depend on polysome integrity, as described in the cases of stress granules (SG) and processing bodies (PB) (8 -10). We found that murine Smaug 1 is expressed in the brain and is abundant in synaptoneurosomes, a subcellular region where translation is tightly regulated by synaptic stimulation (reviewed in Refs. [11][12][13][14][15][16][17][18]. Our results suggest a role for Smaug 1 in RNA granule formation and translation regulation of SRE-containing transcripts at postsynaptic sites.

EXPERIMENTAL PROCEDURES
Plasmids and Library Screening-A pCDNA3.0 vector (Invitrogen) encoding Drosophila Smaug was generated by subcloning the coding region from a dSmaug cDNA kindly provided by Dr. C. Smibert (University of Toronto, Canada) using the primers 5Ј-TAAGAACTATCC-CGGTACCACAA-3Ј and 5Ј-GATCAAATTTGCTCGAGTTCTCC-3Ј. Firefly luciferase reporters carrying three copies of either wild type or mutated SRE were constructed by subcloning of the BamHI/HindIII fragment of C145 and C146 plasmids, a generous gift of C. Smibert (19), into a pcDNA3.0 vector. A pCDNA6.0 encoding murine Staufen 1 (GenBank TM accession number AF395842) (20) was used. The predicted coding region of hSmaug 1 from the AK034323 EST was subcloned between HindIII and SacII sites in the pECFP-N1 vector (Clontech) and KpnI and XhoI sites in the pcDNA6.0 vector (Invitrogen).
Primary Antibody against Mammalian Smaug 1-Rabbit polyclonal antisera were raised against the hSmaug 1 SAM domain, which was prepared as similarly described for the Drosophila SAM domain (21).
Translation and Real Time PCR Assays-The following three plasmids were cotransfected into BHK-21 cells plated in 60-mm dishes (10 6 cells per dish) using Lipofectamine 2000 (Invitrogen) in Dulbecco's modified Eagle's medium (Sigma) without serum and antibiotics: 1) 20 -500 ng of pCDNA3.0 vector carrying the firefly-luciferase coding region fused to a SRE tandem repeat, either wild type or mutated (2); 2) 10 g of pECFP-hSmaug 1 or 6.7 g of pECFP vector (Clontech); and 3) 50 ng of pRL-CMV (Promega, Madison WI) encoding Renilla luciferase. Twenty four hours after transfection, firefly and Renilla luciferase activities were quantified using the dual-luciferase reporter assay system (Promega). Cell lysis was performed as indicated by the manufacturers, and measurements were both taken immediately and after freeze-thawing, with no significant variations. Transfections were performed in triplicate, and results were expressed as the average ratio of firefly to Renilla luciferase activity.
Real time RT-PCR using SYBR Green (Molecular Probes, Eugene, OR) was performed in triplicates of 1:10 serial dilutions of RNA using the following primers: 5Ј-tcgcggttgttacttgactg-3Ј and 5Ј-cgatcttgttacaacacccc-3Ј for firefly and 5Ј-ggaaacggatgataactggtcc-3Ј and 5Ј-aggccgcgttaccatgta-3Ј for Renilla luciferase. Following Equation 2 in Ref. 22, cycle threshold (CT) versus log [RNA dilution] was plotted. Only dilutions where plot slopes (ϭ 1 ϩ E (where E is the efficiency of target amplification), according to Ref. 22, were close to 2 and similar for the two templates (difference lower than 5%) were considered, indicating similar amplification efficiencies (22). A linear regression was applied to calculate the ⌬CT (firefly minus Renilla) and the corresponding S.D. The ratio of firefly luciferase mRNA to Renilla luciferase mRNA was calculated as (1 ϩ E) ⌬CT , where (1 ϩ E) was calculated from the slope, and the standard deviation was calculated by error propagation.
Cycloheximide and puromycin (Sigma) were added to conditioned media from stock aqueous solutions. Treatment with 50 ng/ml leptomycin B (Sigma) was performed in conditioned media for 2 h.

Mammalian Smaug 1 Represses Translation of SRE-containing mRNAs-
The two genes homologous to Drosophila Smaug present in mammalian genomes were named as Smaug 1, located in human and mouse chromosome 14, and Smaug 2, located in human chromosome 19 and mouse chromosome 7 (Fig. 1A). We performed a screening of mouse brain, liver, testis, and whole embryo libraries and found two mSmaug 1 transcripts. The most frequent of them corresponds to a previously reported EST (GenBank TM accession number AK034323), and the second most frequent corresponds to a novel late embryonic 5Ј-untranslated region splicing variant (Fig. 1B). No splicing variants involving the coding region were found in our screening nor in the available data bases, including human and mouse sequences. We confirmed the presence of mSmaug 1 transcripts in brain, kidney, heart, and liver by RT-PCR analysis (Fig. 1C).
We sought to investigate the effect of mammalian Smaug 1 on translation and stability of SRE-containing messengers. Although the direct interaction of hSmaug 1 with SRE-containing messengers was not addressed in this study nor in previous reports, it has been shown previously that SAM domains of vertebrate Smaug homologues specifically recognize SREs (2,5). BHK cells were cotransfected with plasmids encoding an enhanced cyan fluorescence protein (ECFP) or a hSmaug 1-ECFP chimera together with firefly luciferase-translational reporters bearing either wild type or mutated SREs in a trimeric tandem array (3xSREϩ and 3xSREϪ, respectively) (19). A plasmid encoding Renilla luciferase to normalize for transfection efficiency was cotransfected. We found that in the presence of hSmaug 1-ECFP the firefly luciferase activity yielded from 3xSREϩ-containing transcripts was reduced to 43% relative to the ECFP control. In contrast, firefly luciferase activity yielded from 3xSREϪ reporters was the same in the presence of hSmaug 1-ECFP or ECFP ( Fig. 2A). When extracts of cells expressing firefly luciferase were preincubated with extracts of cells expressing hSmaug 1-ECFP, no effects on luciferase activity levels were observed (not shown), further indicating that luciferase reporters are not modulated at the protein level by hSmaug 1.
Simultaneously, firefly and Renilla luciferase mRNA levels were determined by real time PCR. The abundance of the firefly reporters relative to the coexpressed control Renilla mRNA was not affected by the expression of hSmaug 1 (Fig. 2B). As expected, the translational repression was proportional to the relative amounts of hSmaug 1 (Fig. 2C).
These results indicate that human Smaug 1 represses translation of SRE-containing messengers without affecting their stability. The possibility of an effect on mRNA decay or deadenylation by recruitment of the CCR4 deadenylase, as is the case of the Drosophila molecule (7), in a different molecular or cell context was not analyzed, and thus this remains open.
Murine Smaug 1 Is Present in Brain Synaptoneurosomes and Forms Small Particles-An antibody raised in our laboratory against the native SAM domain of hSmaug 1 specifically recognized the human protein in Western blot assays of BHK cells expressing hSmaug 1-V5. Expression of endogenous Smaug 1 was not detected in BHK or other fibroblasts cell lines (Fig. 3A and data not shown).
Next, we evaluated the expression of Smaug 1 in the brain, as translational regulation is frequent in neurons (11, 13-15, 17, 24). A protein band of ϳ70 kDa was observed in blots of mouse and rat brain extracts, in accordance with the predicted molecular weight of mSmaug 1 (Fig.  3B). We then examined the distribution of mSmaug 1 in brain subcellular fractions. Murine Smaug 1 was enriched in synaptoneurosomes (Fig. 3C), and a further purification indicated that mSmaug 1 concen-   DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 trates in post-synaptic densities, a fraction from synatoneurosomes almost membrane-free and enriched in cortical cytoskeleton and synapse-associated elements, such as localized mRNAs and polyribosomes and the marker protein PSD-95 (11,13,16). In contrast, mSmaug 1 was . Smaug 1 is present in brain synaptoneurosomes as 20 S particles. A, polyclonal antibody against the SAM domain of the human Smaug 1 molecule (hSmg1) was used in Western blot assays. Extracts from BHK cells transfected with a hSmaug 1 construct fused to V5 (ϩ) or nontransfected (Ϫ) were analyzed. Equivalent amounts of protein extracts were loaded, and ␤-actin was detected as a loading control. Dilutions of the primary antibody are indicated. No signal was observed with a preimmune serum. B, Western blot analysis of adult mouse (Mo) and P0 rat brain. C, Western blot analysis (left) of total brain (B), crude synaptoneurosomes (S), membrane from synaptoneurosomes (M), postsynaptic densities (P), and rat oligodendrocyte primary culture extract (O). Right, Coomassie Brilliant Blue staining of the same filter. Twenty g of protein were loaded in total brain, crude synaptoneurosomes, and rat oligodendrocyte primary culture extract, and 5 g in membrane from synaptoneurosomes and postsynaptic densities. D, PSD extract was analyzed in sedimentation sucrose gradients (20 -60%, top panel or 10 -30%, bottom panel). Individual fractions were collected, and the presence of mSmaug 1, Staufen 1 (Stau1), and the small ribosomal marker S6 were simultaneously detected by Western blot. ␤-galactosidase was included in the bottom gradient as a 20 S marker. not detected in the synatoneurosomes membrane fraction (Fig. 3C). It also seems to be absent from myelin, as an oligodendrocyte extract gave no signal (Fig. 3C). Taken together, these results suggest that mSmaug 1 is localized in neuronal structures associated with dendritic synapses.

Translation Repressor Smaug Forms Stress Granule-like Foci
Next, we investigated the interaction of mSmaug 1 with the synaptoneurosomal translational apparatus by sedimentation velocity analysis. After centrifugation through a sucrose gradient, mSmaug 1 migrated as a 20 S particle and appeared not to be associated with polyribosomes or with ribosomal subunits (Fig. 3D). As expected, the double-stranded RNA-binding protein Staufen 1 was detected in polysomes and in a faster sedimenting fraction (Fig. 3D), as described previously (16,20,(25)(26)(27)(28). These results are compatible with a role of mSmaug 1 in inhibition of translation initiation in neurons.
Smaug Forms Cytosolic Granules That Contain Polyadenylated RNA-To analyze further translational repression by mammalian Smaug 1, tagged hSmaug 1 molecules were transiently transfected in fibroblast cell lines, and their subcellular distribution was then analyzed by immunofluorescence and sedimentation velocity gradients.
We found that a large proportion of Smaug 1 concentrated in cytoplasmic granules no larger than 0.5-2 m (Fig. 4A). These granules were observed in COS-7, BHK, and HeLa cells transfected with two different human Smaug 1 chimeras, fused to ECFP or to the small tag His 6 -V5 ( Fig. 4 and data not shown). Most cells (66 -85%) showed the representative phenotype depicted in Fig. 4A, granule size, and number as being roughly proportional to expression levels (TABLE ONE). As expected, transfected ECFP presented no granules (data not shown). Human Smaug 1 granules were distributed throughout the cytoplasm and were also observed after Triton X-100 extraction of living cells (Fig.  4B), whereas the nonpunctate hSmaug 1 signal was no longer present (see also TABLE ONE). Treatment with the CRM1-inhibitory drug leptomycin B strikingly increased the otherwise minor nuclear hSmaug 1 staining (Fig. 4, B and C), suggesting that this protein is shuttling between the nucleus and the cytoplasm. Smaug 1 foci were extremely infrequent in the nuclear compartment; the protein showed a homogeneous distribution throughout the nucleoplasm (Fig. 4C).
Remarkably, Drosophila Smaug also formed cytoplasmic granules when expressed in fibroblast cell lines (Fig. 4D). Moreover, the Drosophila protein colocalized with mammalian Smaug 1 when coexpressed (Fig. 4E), further suggesting a functional similarity between the fly and the mammalian homologues.
We then performed a subcellular fractionation of BHK cells expressing hSmaug 1-V5. Transfected hSmaug 1 was quantitatively recovered in the cytosolic fraction. When post-nuclear extracts were analyzed by sedimentation velocity centrifugation, hSmaug 1 migrated close to the 20 S marker (Fig. 4F), similar to the synaptoneurosome-associated mSmaug 1 (Fig. 3D). We simultaneously analyzed the distribution of the Drosophila protein coexpressed in the same conditions. Most interestingly, fly Smaug also migrates as a small ribonucleoparticle or free protein (Fig. 4F). Finally, for comparison, a construct encoding Staufen 1-V5 was cotransfected with hSmaug 1-V5 and analyzed simultaneously. Similarly to the endogenous Staufen 1 (Fig. 3D), Staufen 1-V5 comigrated with polysomes (Fig. 4F), indicating that overexpression is not perturbing the subcellular distribution of the transfected proteins. Furthermore, the absence of the transfected dSmaug and hSmaug 1 in fast sedimentation fractions, together with the plasticity of hSmaug 1 foci (see below), indicate that Smaug foci were not aggregates of misfolded proteins.
Given that hSmaug 1 repressed the translation of SRE-containing transcripts (Fig. 2), we investigated the presence of messenger RNA in the hSmaug 1 foci. Detection of the general PABP in these structures indicated that Smaug granules likely contain polyadenylated RNA. Remarkably, PABP became granular in all cells bearing hSmaug 1 foci, regardless of their size (Fig. 5, A, B, and D). This was observed in both BHK and COS-7 cells at different expression times and by using polyclonal or monoclonal anti-PABP antibodies (Fig. 5, A and B). PABP presented a uniform cytoplasmic staining in nontransfected cells (Fig.  5A), and expression of ECFP had no effect on PABP distribution (Fig.  5C). PABP was retained in hSmaug 1 foci after Triton X-100 extraction (Fig. 5D), and strict colocalization was confirmed in all foci in 94% of the cells, regardless of size and subcellular localization (Fig. 5D, see also

Presence of TIA-1, TIAR, and PABP in hSmaug 1 foci
The frequency of hSmaug 1 foci and the presence of marker proteins in those foci were analyzed in independent transient-transfection experiments by immunostaining and confocal microscopy. Determinations were performed after 16 h of expression, unless indicated otherwise. N indicates number of transfected cells analyzed. Cells with more than four hSmaug1 granules larger than 0.5 m were considered as containing foci. The percentage relative to total transfected cells is indicated. The percentage of cells showing colocalization with the different markers was relative to the number of granule-containing cells. Cells were scored as positive for colocalization when more than 60% of their hSmaug1 foci contained the analyzed marker. Cells positive for PABP colocalization always showed more than 95% of double-stained foci. No differences in colocalization frequency of the distinct markers were observed between high expressing and low expressing cells (not shown). ND indicates not determined.

Construction
Cell  DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52  (26,29), was also compared. Human Smaug 1 and the endogenous SMN localized in distinct cytoplasmic foci, whereas Staufen 1 was frequently detected in hSmaug 1 granules (Fig. 5E). Similar to PABP, Staufen 1 distribution was not affected by ECFP expression, indicating that PABP and Staufen 1 were selectively recruited to hSmaug 1 foci. Cytoplasmic RNA granules are compatible with degradation foci, which are ubiquitous structures containing proteins involved in RNA decay (30). Thus, we investigated whether GW182, an RNA-binding protein marker of degradation foci (31,32), was present in hSmaug 1 granules. Complete reciprocal exclusion of GW182 and hSmaug 1 was observed in the foci in two independent experiments in HeLa cells. The overall distribution of GW182 remained unchanged in hSmaug 1-expressing cells (Fig. 5F).

Translation Repressor Smaug Forms Stress Granule-like Foci
We concluded that in fibroblast cell lines both mammalian hSmaug 1 and fly Smaug form 20 S particles that do not contain ribosomal subunits and that cluster into larger granules containing polyadenylated RNA. In addition, cytoplasmic hSmaug 1 granules do not overlap with the previously described degradation foci or SMN granules.

Smaug 1 Foci Are in Equilibrium with Polysomes and Contain SG
Markers-As our results indicate that Smaug 1 granules may contain polyadenylated RNA in a silent state, and cytoplasmic granules containing silenced mRNA can be in dynamic equilibrium with translating polysomes (8,16,30,33), we sought to investigate the effect of drugs that affect polysome stability on hSmaug 1 foci. We found that after a strong treatment with the polysome-stabilizing drug cycloheximide, hSmaug 1 granules disappeared almost completely (Fig. 6). The effect was timeand dose-dependent; both granule size and number were always smaller after cycloheximide exposure (Fig. 6, B and C, and data not shown). Remarkably, puromycin, which inhibits translation by disrupting polysomes, provoked the opposite effect; more hSmaug 1 granules were observed, and these granules were significantly larger (Fig. 6, B and C). Inhibition of protein synthesis by either cycloheximide or puromycin did not affect hSmaug 1 levels (Fig. 6A). The disruption of hSmaug 1 granules by cycloheximide treatment together with the granule-enhancement effect elicited by puromycin indicate that hSmaug 1 foci are dynamic and that their formation and dissolution are linked to polysome integrity.
Smaug 1 foci plasticity is reminiscent to that of SGs that are induced during cell stress and contain transiently silenced mRNAs (8). Remarkably, we found that Staufen 1, which was recently reported as an SG component (20), was recruited into hSmaug 1 foci (Fig. 5E). Therefore, we investigated the presence of SG markers in hSmaug 1 foci. We focused on TIAR and TIA-1, two nucleus-cytoplasm shuttling proteins that are present in SG (8,33,34). Most hSmaug 1 foci were observed to contain TIAR and TIA-1 in independent transfection experiments performed in BHK or COS-7 cells (Figs. 7 and 8). Colocalization with TIA-1 and TIAR was observed in both the small foci (Fig. 7A) as well as in the large foci present at higher expression levels (Fig. 7B) and resisted the extraction with nonionic detergents (Fig. 7C and TABLE ONE). As reported previously (35), the presence of cytoplasmic granules containing TIAR/TIA-1 was observed in 15-20% of ECFP-transfected cells and was less than 5% in nontransfected cells (data not shown), indicating a minor cellular stress likely due to plasmid transfection and general protein overexpression. HuR, another nuclear protein present in stress granules (34), was also frequently detected in the hSmaug 1 foci (Fig. 8).
In both extracted and nonextracted cells, we observed that about 25% of cells expressing hSmaug 1-ECFP or hSmaug 1-V5 lacked TIAR and TIA in some foci ( Fig. 7C and TABLE ONE). A similar phenomenon was observed previously for G3BP-induced stress granules (35). We found that these two SG-marker proteins were simultaneously present or absent from hSmaug 1 foci. More importantly, the TIAR/TIA-1-negative foci always contained PABP (Fig. 7, C and D), thus suggesting that the presence of polyadenylated transcripts was independent of the recruitment of TIAR/TIA-1.
The two previously described foci, SG and PB, were recently shown to interact physically, displaying encapsulation of PB by SG, juxtaposition, or fusion (30,36,37). The spatial relationship between hSmaug 1 foci and SG markers was analyzed by confocal slicing (Fig. 8). The intensity profile of the hSmaug 1 signal along a line crossing several foci strictly correlated with that of TIAR (Fig. 8A) and HuR (Fig. 8B), indicating that these molecules overlapped spatially in the foci. In contrast, the PB marker GW182 did not colocalize with hSmaug 1, although they were frequently observed in adjacent foci ( Fig. 5F and data not shown).
Altogether, these observations indicate the following: transiently transfected hSmaug 1 forms foci distinct from processing bodies; are in equilibrium with translating polysomes; always contain polyadenylated RNA, and frequently contain stress-granule markers.

DISCUSSION
We have characterized Smaug 1, a mammalian homologue of Drosophila Smaug, an RNA-binding protein defining a novel family of posttranscriptional regulators involved in mRNA silencing and deadenylation (2)(3)(4)7). We found that hSmaug 1 represses the translation of SRE-containing messengers and no effect of human Smaug 1 on mRNA stability was observed in fibroblast cell lines. It is still possible that hSmaug 1 or other mammalian orthologues may have a role in RNA degradation in a different cell context. We found that in the absence of hSmaug 1 constructs, the abundance of SRE-containing transcripts was lower than that of transcripts bearing mutated SRE, and simultaneously, the translation efficiency was stimulated by the presence of SRE motifs (data not shown). This likely suggests the presence of unknown endogenous factors regulating the activation or stability of SRE-containing transcripts that was not analyzed further.
The mechanism involved in mRNA repression by mammalian Smaug 1 has not been addressed in this study, but it is likely to be similar to the one reported for Drosophila Smaug; dSmaug binds Cup, which in turn  DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 binds to eIF4E, thus preventing eIF4G recruitment (6). As a consequence, translation initiation is blocked, and ribosome subunits cannot be recruited. A human Cup homologous protein has been identified, and similar to Drosophila Cup, this ubiquitous molecule, known as 4ET, inhibits the interaction of eIF4E with eIF4G (39). Our observation that Drosophila and mammalian Smaug form 20 S particles that do not include 40 S ribosomal subunits is compatible with a role for these molecules in translation initiation blockage. These 20 S particles appear to build up large aggregates inside cells, which can be visualized as granules. The formation of hSmaug 1 cytoplasmic granules was observed in several different cell lines, and in all cases, they contained PABP, suggesting the presence of polyadenylated SRE-containing messengers. We found that hSmaug 1 foci enlarge when polysomes are disrupted, and conversely, they disassemble in the presence of the polysome-stabilizing drug cycloheximide, as reported for stress granules and more recently for processing bodies (8,16,30,33,37).

Translation Repressor Smaug Forms Stress Granule-like Foci
In addition to PABP, hSmaug 1 foci frequently contained TIAR, TIA-1, and HuR, three RNA-binding proteins that are normally concentrated in the nucleus and that are recruited to cytoplasmic stress granules (8 -10, 34) and Staufen 1, a ubiquitous double-stranded RNAbinding protein recently described to be an SG component (20).
The finding that both fly and mammalian Smaug induce the formation of cytoplasmic granules containing stress granule markers in the absence of cell stressors was unexpected. Nevertheless, this observation is compatible with the fact that dSmaug and hSmaug 1 are translational repressors, given that stress granules are formed when translation initiation is impaired by cellular stress, expression of a phosphomimetic mutant of eIF2␣, or translation inhibitory drugs (8,9,20,33,38,40). Furthermore, stalled translation initiation complexes have been suggested to function as nucleation centers for SG aggregation (10).
By extending our observations on Smaug, we speculate that SG-related structures would assemble onto core complexes of mRNA and proteins that block translation initiation. Relevantly, the translation inhibitor CPEB was recently shown to form similar granules in the absence of stress when overexpressed (36). Despite the presence of SG markers in hSmaug 1 cytoplasmic foci observed under normal conditions, these foci do not necessarily represent structures identical to the stress granules observed upon induction of cellular stress, as SG formation is a consequence of the activation of a complex signal transduction pathway not expected to be activated by Smaug expression.
As an alternative to the de novo formation of granules induced by Smaug, we should consider the possibility of Smaug being recruited to pre-existing cytoplasmic structures, like the PB that are present in all cell type (30,31,32,36,37). Relevantly, fly Smaug and the yeast homologue Vts1 trigger mRNA deadenylation by direct interaction with CCR4 (2, 7) that localize at degradation foci in mammalian cells (30). In addition, 4ET, a mammalian homologue of Cup, a molecular partner of Drosophila Smaug, was recently reported to be present in processing bodies (37,41). We found that Smaug 1 foci contain PABP in BHK, COS-7, and HeLa cells, likely indicating the presence of polyadenylated mRNA, which is normally excluded from degradation foci (30,37) but that can be detected when mRNA degradation is impaired (30). Nevertheless, we found that the presence of hSmaug 1 and the degradation foci marker GW182 in cytoplasmic granules are mutually exclusive in HeLa cells.
Thus, Smaug 1 granules containing silenced mRNAs may be expected to be present at the postsynapse, where this protein accumulates. The formation of hSmaug 1 cytoplasmic granules in fibroblasts indicates that neuron-specific factors are not required, thus underscoring the capacity of Smaug as an RNA granule-forming protein and suggesting the presence of similar structures in neurons. Relevantly, Drosophila Smaug also concentrates in granules, likely the polar granules, at the posterior of the embryo and is present in the fly central nervous system during embryogenesis (3,4). Thus, it seems possible that Smaug 1 molecules located at post-synaptic sites specifically sequester SRE-containing transcripts in granules, where translation is suppressed. More importantly, a growing number of RNA-binding proteins thought to be involved in the local regulation of gene expression are present in dendrites and post-synaptic sites forming granules (11,26,29,(42)(43)(44)(45). Relevantly, some of them, namely Staufen, SMN, fragile X mental retardation protein, and ELAV family members, are also found, under certain conditions, in larger accretions eventually identified as stress granules in a number of cases (20, 46 -48).
Mechanisms for translational regulation contributing to synapse for-  DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 mation and plasticity include CPEB-mediated poly(A) elongation (15), use of an internal ribosome entry site (49,50), and translation regulation by fragile X mental retardation protein and BC1 (24), among others. In addition, availability and activity of general factors for protein synthesis are also regulated at the synapse (11,14,17,18). Our results suggest the SRE-Smaug system as a novel additional mechanism for local translational control at the synapse. A survey of neuronal messengers containing SREs in data bases yielded a reduced number of mRNAs encoding distinct functions. 4 Some of them are known to be localized at the dendritic compartment, and thus, the possibility exists that Smaug contributes to the regulation of these transcripts. This hypothesis and whether Smaug is controlled by synaptic activity have yet to be confirmed.