PKCα Binds G3BP2 and Regulates Stress Granule Formation Following Cellular Stress

Protein kinase C (PKC) isoforms regulate a number of processes crucial for the fate of a cell. In this study we identify previously unrecognized interaction partners of PKCα and a novel role for PKCα in the regulation of stress granule formation during cellular stress. Three RNA-binding proteins, cytoplasmic poly(A)+ binding protein (PABPC1), IGF-II mRNA binding protein 3 (IGF2BP3), and RasGAP binding protein 2 (G3BP2) all co-precipitate with PKCα. RNase treatment abolished the association with IGF2BP3 and PABPC1 whereas the PKCα-G3BP2 interaction was largely resistant to this. Furthermore, interactions between recombinant PKCα and G3BP2 indicated that the interaction is direct and PKCα can phosphorylate G3BP2 in vitro. The binding is mediated via the regulatory domain of PKCα and the C-terminal RNA-binding domain of G3BP2. Both proteins relocate to and co-localize in stress granules, but not to P-bodies, when cells are subjected to stress. Heat shock-induced stress granule assembly and phosphorylation of eIF2α are suppressed following downregulation of PKCα by siRNA. In conclusion this study identifies novel interaction partners of PKCα and a novel role for PKCα in regulation of stress granules.


Introduction
Protein kinase C (PKC) is a family of serine/threonine kinases that play important roles in several processes that control cell fate such as apoptosis, proliferation, and differentiation. The PKC isoforms are grouped in classical (PKCa, bI, bII, and c), novel (PKCd, e, g, and h) and atypical (PKCf, and i) PKCs [1]. Each isoform can be specifically regulated and has unique functions in a given cell. This is conceivably in part attributed to differential sensitivities to activating factors that are increased upon stimulation of cell surface receptors. For example, only classical isoforms are activated by Ca 2+ and only classical and novel isoforms are sensitive to diacylglycerol. In addition to these differences in activator sensitivity, a wide range of studies indicate that isoformspecific interactions with other proteins largely determine the function of a PKC isoform [2].
In order to further understand PKC isoform function we have screened for interaction partners by immunoprecipitating a neuritogenic PKCe structure and identified co-precipitated proteins by mass spectrometry analysis. One of these was the intermediate filament protein peripherin [3]. In addition, several proteins containing RNA recognition motifs (RRMs) were identified. These include the cytoplasmic poly(A) + binding protein (PABPC1), the IGF-II mRNA binding protein 3 (IGF2BP3), and the RasGAP binding protein 2 (G3BP2). PABPC1 binds to the poly(A) tail of mRNAs and its primary role is conceivably in regulating translation initiation and mRNA stability [4]. IGF2BP3 was first detected overexpressed in pancreatic cancer and initially called KOC (after KH domain containing protein overexpressed in cancer) but later identified as an IGF-II mRNA-binding protein and re-named [5,6]. It is one of three members of the IGF-II mRNA-binding protein (IMP) family which are important for transporting their mRNA targets to proper cellular localization during development [7]. G3BP2 and the closely related protein G3BP1, which has been more studied, may have several roles in the cell. They have both been shown to interact with SH3 domains in GTPase-activating proteins [8,9] and with MYC mRNA [10] regulating its turnover. Furthermore, overexpression of G3BP1 leads to the assembly of stress granules which sequester and retain mRNA-species that are not supposed to be translated during the cellular stress response [11,12].
When analyzing endogenous proteins we could detect an interaction between PKCa and the three mRNA-binding proteins. The fact that all proteins contain RNA-binding domains, and that at least PABPC1 and G3BP are known to localize to stress granules, led us to raise the hypothesis that PKC may participate in stress granule formation. In this paper we demonstrate that PKCa both regulates the assembly of stress granules and is a component of them. The results therefore provide information regarding novel cellular roles of PKCa.

PKCa interacts with G3BP2
In a screen for proteins that interact with an overexpressed PKCe construct (PKCePSC1V3) [3], mass spectrometry analysis of some of the co-precipitated proteins resulted in the identification of the RNA-binding proteins IGF2BP3, PABPC1, and G3BP2. Although the interactions could be confirmed for overexpressed PKCe constructs we could not detect endogenous interactions under normal growth conditions. However, we found that the proteins interacted with PKCa (Fig. 1).
The fact that the identified PKC-interacting proteins also are RNA-binding proteins raised the possibility that the interaction is dependent on intact RNA. We therefore analysed the interaction in cell lysates that had been treated with either RNase inhibitor or RNase prior to immunoprecipitation (Fig. 1A). The interaction of PKCa with IGF2BP3 and PABPC1 was abolished by RNase treatment whereas G3BP2 and PKCa could still be co-precipitated with each other. G3BP2 itself interacted with IGF2BP3 and PABPC1 in an RNA-dependent manner, which could suggest that the proteins are in a common complex.
To analyze the structures in PKCa that mediate the interactions, EGFP-fusions of PKCa domains were expressed in SK-N-BE(2)C neuroblastoma cells and immunoprecipitated using the EGFP tag (Fig. 1B). G3BP2, PABPC1 and IGF2BP3 all coprecipitated with essentially the same PKCa constructs. The binding seems to be mediated primarily by the regulatory domain, where the C1a and C2 domains both have binding capability. On the other hand, the C1b domain did not interact with the proteins.
The data in Figure 1A indicate that there may be a direct binding between PKCa and G3BP2, whereas the association with IGF2BP3 and PABPC1 is indirect and dependent on RNA. Therefore we further analyzed the PKCa-G3BP2 interaction with a GST pull-down experiment with recombinant proteins (Fig. 1C). PKCa was pulled down together with GST-G3BP2 demonstrating a direct interaction between the proteins. Both G3BP2 variants (G3BP2a and G3BP2b which are generated by alternative splicing) bind PKCa. The binding is not specific for G3BP2 since the closely related protein G3BP1 also pulled down PKCa (Fig. 1C).
To examine which structures in G3BP2 that mediate the binding, GST-tagged domains of G3BP2 were included in the binding assay (Fig. 1D). The isolated RNA-binding domain pulled down PKCa whereas G3BP2 variants lacking this domain did not interact.
Taken together, the data suggest that the C1a and C2 domains in the regulatory domain of PKCa contain structures that can mediate the interaction with the C-terminal RNA binding domain of G3BP2.
We also analyzed if G3BP2 is a PKC substrate in vitro ( Fig. 1E  and 1F). G3BP2 was phosphorylated in the presence of PKCa, although not to the same extent as PKCa itself. The phosphorylation took place also in the absence of PKC activators (Fig. 1E). It is conceivable that the NTF, but not the RRM, domain contains PKC phosphorylation sites since GST-NTF but not GST-RRM was phosphorylated in the kinase assay (Fig. 1F).

A slower migrating PKCa variant is enriched in G3BP2 precipitates
As can be seen in Figure 1A the major PKCa species that is coimmunoprecipitated with G3BP2 does not migrate as fast as the major PKCa species in the PKCa immunoprecipitate. To certify that this is PKCa and to investigate whether there is a difference in post-translational modifications of the PKCa variants, PKCa and G3BP2 immunoprecipitates were probed with different PKCa antibodies (Fig. 2). The upper band, enriched in G3BP2 precipitates, is clearly recognized by an antibody towards the Nterminal region of PKCa but more weakly by an antibody towards the C-terminal region. The antibody recognizing phosphorylated T638 identified both PKCa bands in a ratio similar as the general PKCa antibodies ( Fig. 2A). However an antibody directed towards phosphorylated S657 primarily reacted with the upper band. Both antibodies towards phosphorylated PKC recognized the PKCa found in G3BP2 precipitates (Fig. 2B). Thus, the fact that several PKCa antibodies react with the slow-migrating species in the G3BP2 precipitate underscores that it is PKCa. G3BP2 primarily interacts with a PKCa variant that has both its C-terminal phosphorylation sites phosphorylated.
PKCa as well as G3BP2, IGF2BP3 and PABPC1 localizes to stress granules PABPC1 and G3BP1 (closely related to G3BP2 and often referred to as G3BP) have previously been shown to be associated with the formation of stress granules during cellular stress [11,13]. We therefore investigated whether the three PKC-interacting proteins co-localize during stress. SK-N-BE(2)C cells were incubated at 44uC and PABPC1, G3BP2 and IGF2BP3 proteins were visualized by immunofluorescence (Fig. 3A). Under normal conditions PABPC1, G3BP2 and IGF2BP3 (not shown) are diffusely localized in the cytosol, whereas upon heat shock the proteins relocate to newly formed stress granules. Thus, the proteins co-localize in stress granules under these conditions.
To investigate whether PKCa accompanies the identified mRNA-binding proteins to stress granules SK-N-BE(2)C cells were subjected to heat shock and PKCa and PABPC1 were tandem C1a and C1b (C1ab), or PSC1 (PSC1) domains. The constructs are illustrated below the blot. Cell lysates were immunoprecipitated using anti-GFP-conjugated magnetic beads. Lysates and precipitates were thereafter analyzed with Western blot using IGF2BP3, PABPC1, G3BP2 or GFP antibodies. PKCa was incubated with GST-tagged G3BP1, G3BP2a and G3BP2b (C) or different G3BP2b constructs (D). Thereafter GST was pulled down with GSH-coupled sepharose. Incubation mixture and pull-downs were analyzed with Western blot. Schematic illustration of the G3BP2 constructs is shown below the blot. (E) G3BP2b (with GST cleaved off) was incubated with PKCa, [c-32 P]ATP and PKC activators (Ca 2+ /PS/DAG -Ca 2+ , phosphatidylserine and diacylglycerol) as indicated. The reactions were separated on SDS-PAGE, blotted and [ 32 P] was visualized with autoradiography and proteins with immunoblotting. (F) G3BP2b (with GST cleaved off) and GST-tagged G3BP2 domains were incubated with PKCa, [c-32 P]ATP and PKC activators. The reactions were analyzed as in (E). doi:10.1371/journal.pone.0035820.g001 visualized by immunofluorescence (Fig. 3B). PKCa and PABPC1 both showed a diffuse cytosolic localization pattern in cells cultured at 37uC. After 1 h of heat shock PABPC1-containing stress granules were formed and PKCa was present in many of the PABPC1-containing granules (Fig. 3B). Accumulation of PKCa in PABPC1-containing stress granules was also observed after treatment with As 2 O 3 (Fig. 3B) indicating that the PKCa relocation is not limited to the stress response induced by heat shock. A weak increase of PKCa reactivity in the nucleus could be seen after heat shock (Figs. 3B and 4A) but this was not the case following As 2 O 3 treatment.
Processing-bodies (P-bodies) constitute another class of mRNArich granules that are functionally and spatially linked to stress granules [14,15]. To analyze whether PKCa also localizes to these structures, SK-N-BE(2)C cells were subjected to heat shock. PKCa and the P-body marker Dcp1a were thereafter visualized with immunofluorescence ( Fig. 4A). PKCa could not be detected in Dcp1a-positive structures.
To simultaneously visualize P-bodies, stress granules and PKCa, the cells were transfected with a vector encoding EGFPtagged PKCa and stained for G3BP2 and Dcp1a (Fig. 4B). Stress granules and P-bodies were in many cases localized immediately adjacent to each other with some overlapping pixels in the borders of the structures. However, PKCa displayed a clear co-localization with stress granules whereas isolated P-bodies were PKCanegative.
The PKCaC1a domain associates with stress granule proteins after heat shock Our data indicate that the C1a but not the C1b domain of PKCa contains structures that can mediate its interaction with G3BP2 ( Fig. 1B). To investigate if it also can mediate the association with stress granule components, the PKCaC1a and PKCaC1b domains were expressed in SK-N-BE(2)C cells that were subsequently subjected to heat shock. To enrich for stress granule components we immunoprecipitated the stress granule component TIAR (Fig. 5). As expected, G3BP2 was coprecipitated with TIAR following heat shock, indicating that stress granule components are enriched in the precipitate. The experiment revealed that the PKCaC1a domain, but not the PKCaC1b domain, co-precipitates with TIAR upon heat shock, indicating that the C1a domain can mediate interaction with stress granule components.

Downregulation of PKCa delays stress granule formation
Since PKCa interacts with stress granule components and also localizes to these structures we postulated that PKCa may be involved in the regulation of stress granule formation. To test this hypothesis we aimed at downregulating PKCa with siRNA and study the stress granule induction. Due to difficulties in obtaining substantial knockdown of PKCa in SK-N-BE(2)C cells, we used the breast carcinoma MDA-MB-231 cell line for these experiments. PKCa localizes to stress granules upon heat shock also in MDA-MB-231 cells (data not shown). MDA-MB-231 cells were transfected with three different PKCa siRNA oligonucleotides and were thereafter subjected to a 44uC heat shock ( Fig. 6A and 6B). Western blot demonstrated decreased PKCa levels following transfections with PKCa siRNA. Silencing of PKCa led to a decrease of the amount of cells with stress granules under heat shock (from 86%67% for control to 48%613 for siPKCa(I)-, 41%616% for siPKCa(II)-, and 49%614% for siPKCa(III)transfected cells; Fig. 6A and 6B). PKCa downregulation did not delay the disassembly of the stress granules following a reversal of the temperature to 37uC.
The effects of PKCa downregulation on stress granules formation was compared with knock-down of PKCe (Fig. 6C). Suppression of PKCe levels did not influence the formation of stress granules as PKCa did. PKCa downregulation only suppressed stress granule formation during the initial phase suggesting that absence of PKCa does not abolish stress granules but rather delays their assembly. The As 2 O 3 -induced stress granule assembly was not siginificantly reduced in PKCadownregulated cells, suggesting that the importance of PKCa depends on the stress inducer (Fig. 6D). To analyze whether PKC activity affects stress granule formation we treated cells with the PKC activator 12-Otetradecanoylphorbol-13-acetate (TPA) and/or the inhibitor GF109203X concomitantly with heat shock (Fig. 6E). Neither agent induced stress granules by themselves. However, the heat shock-induced stress granule formation was potentiated by the PKC inhibitor.

Downregulation of PKCa delays heat shock-induced phosphorylation of eIF2a
Translational arrest by phosphorylation of eukaryotic translation initiation factor 2a (eIF2a) is one of the major triggers that induce stress granule formation. We therefore analyzed if downregulation of PKCa also leads to suppression of the heat shockinduced eIF2a phosphorylation ( Fig. 7A and 7B). As for stress granule formation, there was a suppression of the initial phosphorylation of eIF2a whereas after prolonged stress no effect of PKCa downregulation could be discerned. As 2 O 3 exposure also leads to increased phosphorylation of eIF2a. However, contrary to heat shock, we could not detect a suppression of the phosphorylation in PKCa-depleted cells (Fig. 7C and 7D).
To investigate whether downregulation of a kinase upstream of eIF2a could explain the effect of PKCa depletion on heat shockinduced eIF2a phosphorylation we analyzed the levels of protein kinase R (PKR) and heme-regulated inhibitor kinase (HRI), two kinases that may mediate eIF2a phosphorylation during heat shock (Fig. 7E). PKR levels were, if anything, increased in PKCadepleted cells. However, HRI levels were lower in PKCa-depleted cell in all four experiments but the p-value was not below 0.05. As a comparison we included lysates from PKCe-depleted cells. The levels of HRI and PKR were not influenced as much in these cells.
To obtain further insights into PKCe we analyzed whether PKCe co-precipitates with IGF2BP3, PABPC1 and G3BP2 (Fig. 7F). We could not detect PKCe in either precipitate under normal growth conditions. However, following heat shock, PKCe could be discerned in IGF2BP3-and G3BP2 precipitates.

Discussion
Here we report for the first time that PKCa is a component of stress granules and that it associates with RNA-binding proteins G3BP2, IGF2BP3 and PABPC1. These findings provide novel information regarding PKC-mediated regulation of the cellular response to stress.
Since the identified PKCa interaction partners all are RNAbinding proteins it suggests a role for PKCa in RNA regulation. Indeed the interaction with IGF2BP3 and PABPC1 was dependent on intact RNA indicating the central role of RNA for the association. On the other hand, the interaction with G3BP2 was largely resistant to RNase treatment and could also be obtained with isolated proteins in vitro, indicating that it is a direct binding.
It has long been recognized that PKC can influence protein synthesis by acting at the RNA level. Several studies have demonstrated that PKC activation leads to increased stability of mRNA species [16][17][18][19][20][21]. The mechanisms by which PKC achieves this are still largely unknown but the mRNA-binding Hu proteins are one group of potential mediators. PKC regulates the Hu proteins both by increasing their expression levels [22] with subsequent stabilization of target mRNAs and by phosphorylation which influences its shuttling in and out of the nucleus [23,24]. A role for PKC in mRNA regulation is also supported by the identification of PKCbII in messenger ribonucleoprotein com- plexes [25]. Upon activation, PKCbII binds RACK1 in the complex and RACK1 was shown to bind both PABPC1 and G3BP2.
The interaction with RACK1 is also important for PKC to modulate ribosomal subunit joining [26]. Our results show that PKCa interacts with the mRNA binding proteins in cells under basal conditions, and in the case of G3BP2 the interaction is direct, further highlighting that PKCa may have a role in posttranscriptional regulation in general.
Our data particularly support a role for PKCa in the regulation of mRNA that takes place during stress. When cells are exposed to stress, translation is shifted towards synthesis of proteins of importance for the cellular stress response. The translation of other mRNAs is temporarily silenced and they accumulate in stress granules, which contain the small ribosomal subunits, translation initiation factors, and a vast array of RNA-binding proteins [12,27]. The granules are dynamic and as soon as the cell is no longer exposed to stress stimuli they dissolve and protein translation is resumed [13]. Stress granules are formed when stress-sensitive serine/threonine kinases recognize and phosphorylate eIF2a, an important component of the translation initiation complex [28]. Alternatively formation of stress granules can be triggered when translation initiation is blocked at other steps such as inhibition of eIF4 or eIF4G activities or 80S ribosome assembly [29][30][31]. Stress granules can also be induced by overexpression of stress granule components, such as G3BP1 [11], T-cell intracellular antigen-1/T-cell intracellular antigen-related proteins (TIA-1/TIAR) [32], survival of motor neurons protein (SMN) [33], cytoplasmic polyadenylation-binding protein (CPEB) [34] and fragile X mental retardation protein (FMRP/FXR1) [35] in the absence of stress.
It is also becoming increasingly clear that stress granule assembly and disassembly are under the control of a diverse set of proteins that are not directly RNA-binding. Some act on modulating the post-translational modification of stress granulerelated proteins. For example, stress granules are positive for ubiquitin [36] which apparently is of crucial importance since proteasome inhibition leads to stress granule formation [37] and the downregulation of the ubiquitin-binding protein HDAC6 suppresses their formation [36]. A functional screen revealed Olinked N-acetylglucosamine-modified proteins were enriched in stress granules and important for stress granule formation [38]. Phosphorylation of stress granule components also regulates the granules. Focal adhesion kinase (FAK)-mediated phosphorylation of Grb7 leads to its release from stress granules and is accompanied by stress granule disassembly [39] and phosphorylation of G3BP1 suppresses stress granule assembly by inhibiting its oligomerization [11]. Furthermore, the formation is microtubule-dependent [40] and potentiated following knockdown of apoptosis-inducing factors [41].
Our results, that PKCa relocates to these granules during stress and that knockdown of PKCa in MDA-MB-231 cells affects granule assembly after heat shock add PKCa to the list of stress granule regulators. PKCa depletion primarily led to a delay in the assembly which is analogous to the effect caused by depletion of importin a1 or by interference with microtubules [42]. PKCa did not localize to P-bodies demonstrating that it is not associated with all mRNA containing granules but is more specifically involved in stress granule dynamics. The fact that simultaneous incubation with a PKC inhibitor did not suppress stress granule formation indicates that PKCa kinase activity is not directly involved in the pathway leading to stress granules. PKC isoforms have in other system been shown to exert effects independently of its kinase activity [43,44] and this may be another process regulated by PKC in a similar manner. Another alternative explanation could be that lower PKCa amounts during a longer time period may alter levels or functions of components important for stress granule formation. It is conceivable that the effects of PKCa depletion at least partially can be explained by alterations upstream of eIF2a phosphorylation since this event was also delayed in PKCadownregulated cells. It is possible that lower levels of the upstream eIF2a kinase HRI is responsible for the suppressed heat shockinduced phosphorylation of eIF2a in PKCa-depleted cells.
It is likely that the PKCaC1a domain is one mediator of the interaction since this domain, as opposed to the structurally similar C1b domain, was associated with the RNA-binding proteins and was enriched in TIAR precipitates after heat shock. This supports an interrelation between the PKCa interaction with G3BP2 and its localization to and role in stress granule formation. The C1 domains were originally identified as the binding sites for phorbol esters [45] but a number of studies have emerged showing they can mediate protein interactions that are both PKC isoformspecific [46,47] as well as common for several isoforms or C1 domain-containing proteins [48][49][50].
We could also see that G3BP2 preferably associates with a PKCa variant with a slower migration pattern, suggesting that, Figure 5. The PKCaC1a but not C1b domain is recruited to TIAR-containing complexes upon heat shock. SK-N-BE(2)C neuroblastoma cells were transfected with vectors encoding EGFP alone, or EGFP fused to the PKCa C1a or C1b domain and incubated at 37uC or 44uC for 1 h before harvesting. Cell lysates were sonicated and immunoprecipitated using anti-TIAR. Lysates and precipitates were analyzed with Western blot using antibodies indicated in the figure (A). The intensities of the EGFP fusions were quantified and the ratio of the intensity in precipate to the intensity in the lysates is shown in (B). Data are mean 6 SEM, n = 3. doi:10.1371/journal.pone.0035820.g005 Figure 6. Downregulation of PKCa suppresses stress granule assembly. MDA-MB-231 cells were transfected with three different siRNA oligonucleotides against PKCa, and were thereafter subjected to heat shock at 44uC for 1 h. Stress granules were visualized with immunofluorescence post-translational modification of PKCa is of importance for its association with G3BP2. Our analyses indicate differences in the phosphorylation pattern between the major PKCa variant and the one that is enriched in G3BP2 precipitates. One putative explanation to this difference is that a special conformation of PKCa is favorable for the interaction with G3BP2.
In conclusion the data demonstrate novel PKCa interaction partners which open up for mechanistic explanations of PKCa effects on RNA metabolism and stress granule-mediated regulation of the cellular response to stress. using a G3BP2 antibody (A). Western blot of cell lysates demonstrating PKCa downregulation and quantification of the percentage of cells containing stress granules (B) (mean 6 SEM, n = 3). (C) MDA-MB-231 cells were treated with siRNAs targeting PKCa or PKCe, followed by heat shock for indicated time periods. Cell lysates were analyzed with Western blot demonstrating downregulation of respective isoforms. Stress granules were visualized by G3BP2 immunofluorescence and the percentage of cells with stress granules was quantified (mean 6 SEM, n = 3). (D) MDA-MB-231 cells with downregulated PKCa were treated with 300 mM As 2 O 3 for 30 or 60 minutes. Cell lysates were analyzed with Western blot and the percentage of cells with stress granules, identified by PABPC1 immunofluorescence was quantified (mean 6 SEM, n = 3). (E) MDA-MB-231 cells were treated with 16 nM TPA and/or 2 mM GF109203X (GFX) during heat shock. Stress granule-positive cells were thereafter quantified. * denotes statistically significant (p,0.05) difference compared to control using ANOVA followed by Duncan's multiple range test. doi:10.1371/journal.pone.0035820.g006 Figure 7. Heat shock-induced phosphorylation of eIF2a is delayed in cells with downregulated PKCa. PKCa was downregulated in MDA-MB-231 cells by siRNA prior to subjection to heat shock (A and B) or As 2 O 3 treatment (C and D) for indicated time periods. Lysates were analyzed for phosphorylated eIF2a, total eIF2a, PKCa and actin by Western blot (A and C). The levels were quantified and related to total eIF2a and normalized to values obtained in control cells treated with a control siRNA (B and D) (mean 6 SEM, n = 3). (E) PKCa and PKCe were downregulated in MDA-MB-231 cells and the expression levels of HRI and PKR were analyzed with Western blot. The graphs show quantification of HRI and PKR levels divided by actin levels and normalized to control. Data are mean 6 SEM, n = 4. (F) Lysates from cells that had or had not been subjected to heat shock were immunoprecipitated with antibodies towards IGF2BP3, PABPC1 and G3BP2 or matching isotype controls. The precipitates were thereafter analyzed for the presence of PKCe. * denotes statistically significant (p,0.05) difference compared to control using ANOVA followed by Duncan's multiple range test. doi:10.1371/journal.pone.0035820.g007

Plasmids, antibodies and siRNA oligonucleotides
Expression vectors encoding full-length or isolated domains of human PKCa fused to enhanced green fluorescent protein (EGFP) have been described previously [43,51,52]. G3BP2b, G3BP2a and G3BP1 vectors were constructed by PCR of full-length templates (originally obtained from RZPD Deutsches Ressourcezentrum für Genomforschung GmbH for G3BP2b and cDNA from MDA-MB-231 cells for G3BP2a and G3BP1) introducing restriction enzyme sites adapted for cloning in pET41b vector. Primers used are listed in Table 1. All constructs were sequenced.
Sequences of Stealth TM siRNA oligonucleotides (Invitrogen) are listed in Table 1.

Western blot
Proteins were electrophoretically separated by SDS-PAGE and transferred to a PVDF membrane (Millipore). Membranes were pre-incubated with 5% dried milk in PBS followed by incubation with primary antibodies. Membranes were washed, incubated with horseradish peroxidase-labelled secondary antibody, and immunoreactivity was detected with the SuperSignal system (Biological Industries), enhanced chemiluminescence detection system (GE Healthcare) or SuperSignal (Pierce) as substrate. The chemiluminescence was captured with a charge-coupled device camera (Fujifilm) and intensities were quantified with ImageJ.
GST pull-down assay was performed incubating 80 ng PKCa isozyme (Sigma) with 4 mg of GST-G3BP recombinant proteins in 100 ml binding buffer (20 mM Tris, pH 7.4, 0.1 mM EDTA, 100 mM NaCl, 1 mM DTT) with agitation for 1 hour at 4uC. Thereafter 40 ml mMACS TM anti-GST MicroBeads (Miltenyi Biotec) was added and following 1 hour incubation at 4uC, protein separation was performed on a m Column in a magnetic field of mMACS Separator (Miltenyi Biotec) according to manufacturers protocol. GST pull-downs were analyzed with SDS-PAGE and Western blotting.

In vitro kinase assay
Since the GST/His fusion of G3BP2b has the same size as PKCa the GST/His tag was proteolytically removed with thrombin to enable identification on autoradiography. G3BP2b or GST/His fusion of G3BP2 domains (1 mg) was incubated in with 400 ng PKCa (Sigma-Aldrich), 100 mM (2 Ci/mmol) [c-32 P]ATP (Perkin Elmer), 20 mM HEPES (pH 7.4) and 10 mM MgCl 2 . Reactions were either supplemented with 0.5 mM EGTA (absence of activators) or with 0.3% Triton X-100, 100 mg/ml phosphatidylserine (Sigma Aldrich), 0.1 mM CaCl 2 and 20 mg/ml 1,2-diacylglycerol (Avanti). The total volume was 50 ml. Reactions were incubated at 30uC for 20 min and terminated by addition of sample buffer. Samples were separated by SDS-PAGE and subjected to autoradiography and Western blot.

Immunofluorescence and confocal microscopy
Cells were washed in PBS, fixed with 4% paraformaldehyde in PBS for 4 minutes, washed twice in PBS and thereafter permeabilized and blocked with 5% goat serum or 5% bovine serum albumin and 0.3% Triton X-100 in PBS for 30 minutes. Cells were incubated with primary antibodies for 1 h. Following washes in PBS, cells were incubated with secondary Alexa Fluor 488-, 546-, and/or 633-conjugated antibodies in PBS for 1 h followed by extensive washes in PBS and mounting on object slides using 20 ml PVA-DABCO (9.6% polyvinyl alcohol, 24% glycerol, and 2.5% 1,4-diazabicyclo[2.2.2]octane in 67 mM Tris-HCl, pH 8.0).
For confocal microscopy a Bio-Rad Radiance 2000 confocal system fitted on a Nikon microscope with a 60x/NA 1.40 oil lens or a Zeiss LSM710 was used. Excitation wavelengths were 488 nm (EGFP and Alexa Fluor 488), 543 nm (Alexa Fluor 546), and 637 nm (Alexa Fluor 633) and the emission filters used were HQ515/30 (EGFP and Alexa Fluor 488) and 600LP (Alexa Fluor 546). In triple stainings a HQ600/50 bandpass filter was used for Alexa Fluor 546 detection and a 660LP filter for Alexa Fluor 633. For quantification of stress granules 200 cells were scored for the presence of stress granules, identified either by PABPC1 or G3BP2 antibodies.