COS‐7 cells and HeLa cells were cultured as described previously [12]. COS‐7 cells were transfected by electroporation at 220 V/950 mF with a Bio‐Rad gene pulser. The transfection of HeLa cells were performed using FuGENE transfection reagent (Roche, Indianapolis) according to the manufacturers’ procedures.

Plasmids encoding GFP epitope‐tagged G3BP (GFP‐G3BP) was kindly provided by Dr. Jamal Tazi. The pSG5 expression plasmids containing full‐length Tudor‐SN tagged with Flag epitope (pSG5‐Tudor‐SN), the pGEXT‐4T‐1 plasmids containing SN domain (GST‐SN, 1‐639aa), TSN domain (GST‐TSN, 640–885aa) or Tudor domain (678–769aa, GST‐TD) of Tudor‐SN protein were generated as previously described [11, 12]. The full‐length Tudor‐SN (pRFP‐Tudor‐SN), SN (pRFP‐SN, 1–639aa) or TSN (pRFP‐TSN, 640–885aa) fragment was cloned and inserted into the EcoRI/XhoI sites of the vector pCherry‐C1which was kindly provided by Dr. Johan Peranen. All PCR products were sequenced.

GST (glutathione S‐transferase) pull down experiments were performed as previously described [12]. The beads‐bound GST fusion proteins were incubated with the total cell lysate of transfected COS7 cells or in vitro translated ³⁵S‐labeled G3BP protein. After washing, the bound proteins were separated by SDS–PAGE and analyzed by immunoblotting with mouse monoclonal anti‐GFP antibody (Sigma, St. Louis, MO, USA) or autoradiography.

The cell‐free in vitro translation of full‐length G3BP was carried out in a nuclease‐treated rabbit reticulocyte lysate (RRL) system (Promega BioSciences, CA, USA) according to the manufacturer's recommendations. The proteins were labeled with L‐[³⁵S]‐methionine (Amersham Biosciences, USA).

The total cell lysates of HeLa cells without stress stimuli were collected with Nonidet P‐40 lysis buffer (50 mM Tris–HCl, pH 7.6, 300 mM NaCl, 0.1 mM EDTA, 0.5% Nonidet P‐40, 20% glycerol, 0.1 mM sodium orthovanadate, 1 mM sodium butyrate), and then incubated with mouse monoclonal anti‐Tudor‐SN or anti‐G3BP (Abcam, Cambridge, UK), as well as rabbit polyclonal IgG (Santa Cruz biotechnology) as control, followed by incubation with protein‐G/A‐Sepharose (Amersham Pharmacia Biotech). The bound proteins were analyzed by SDS–PAGE and blotted with anti‐Tudor‐SN or anti‐G3BP antibody. The mouse monoclonal anti‐Tudor‐SN antibody was generated against SN4 domain (amino acids 507–674) of Tudor‐SN in Dr. Silvennoinen's lab. The rabbit polyclonal anti‐Tudor‐SN antibody was generated against TSN domain (amino acids 640–885) of Tudor‐SN in our lab.

Cells were grown on glass cover slips. Cellular stress was induced either by treatment with 0.5 mM sodium arsenite (Sigma–Aldrich, St. Louis, MO, USA), or by incubation at 45 °C for different time point (Heat shock). Control cells and treated cells were fixed and permeabilized, and then incubated with mouse monoclonal anti‐G3BP and rabbit polyclonal anti‐Tudor‐SN antibodies, or mouse monoclonal anti‐Tudor‐SN and rabbit polyclonal anti‐TIA‐1 (Santa Cruz biotechnology). After washing, cells were incubated with anti‐mouse Alexa fluor 488 (Invitrogen) and anti‐rabbit Texas‐red (Molecular probes, Eugene, Oregan USA) conjugated secondary antibodies. Confocal images were collected using LSM5 program and Zeiss confocal microscope, equipped with an Argon laser (488 nm) and HeNe laser (543 nm) and a ×63 objective. Green emission was detected using a 505‐nm low pass filter and red emission using a 630‐nm low pass filter [11]. Approximately 200 cells were scored per experiment independently by two different individuals.

For living cell imaging, HeLa cells were transfected with GFP‐tagged G3BP, and RFP‐Tudor‐SN, RFP‐SN or RFP‐TSN, respectively by using FuGENE transfection reagent according to the manufacturers’ procedures. As a control, HeLa cells were transfected with empty vector pEGFP‐C1 and p Cherry‐C1. After 24 h, the cells were seeded onto glass‐bottom dishes (Mat‐Tek, Ashland, MA) and cultured overnight. Before observation, the cells with 2 ml culture medium were maintained in a chamber system at 37 °C and 5% CO₂. The images of timed series were acquired as described above.

HeLa cells were transfected with Tudor‐SN siRNA or scramble siRNA according to previously described [11]. After 72 h, the cells were seeded onto glass‐bottom dishes and cultured overnight. After heat shock at 45 °C for 60 min or treated with 0.5 mM sodium arsenite for 45 min, confocal images were obtained as described above.

Cell proliferation was measured with MTS assay. Briefly, cells were plated in 96‐well plates at a density of 2 × 10³ per well and incubated for 24 h or 48 h, and then the cells were incubated with 20 μl of MTS solution (Promega) for 4 h at 37 °C. The absorbance was measured at 490 nm using ELISA microplate reader Multiskan (Thermo Labsystems).

Tudor‐SN is a multi functional protein composed of four repeats of SN and a Tudor domain followed by a SN5 domain (Fig. 1 A). We initially performed GST‐pull down assay to verify the interaction of Tudor‐SN and G3BP. GST and different GST fusion proteins were bound to glutathione‐coupled beads (Fig. 1B) and incubated with total cell lysates of COS7 cells transfected with GFP‐G3BP. As shown in Fig. 1C, neither Tudor nor TSN domain associated with G3BP, but the SN domain readily precipitated the GFP‐G3BP protein. The beads‐bound different GST fusion proteins were also incubated with in vitro translated ³⁵S‐labelled full‐length G3BP protein. As shown in Fig. 1D, in vitro translated G3BP was found to interact with the GST‐SN fusion protein, but not the others. These results indicate that the SN domain of Tudor‐SN interacts with G3BP.

To substantiate the in vivo interaction of Tudor‐SN and G3BP, the co‐immunoprecipitation experiment was performed with endogenous proteins of HeLa cells. As shown in Fig. 2 A, G3BP only precipitated with anti‐Tudor‐SN antibody, but not the control rabbit IgG antibody. Reciprocally (Fig. 2B), Tudor‐SN was found to co‐precipitate with G3BP, but not the control antibody. These data demonstrate that endogenous Tudor‐SN and G3BP proteins physically form the complex in vivo.

G3BP plays an essential role in SGs formation. To investigate whether Tudor‐SN is also involved in SGs, we examined the localization of endogenous Tudor‐SN and G3BP in response to heat shock for different time points. As shown in Fig. 3 A, in normal HeLa cells, G3BP was distributed in the cytoplasm (green, a), Tudor‐SN was mainly in cytoplasm and little in nucleus (res, b). The co‐localized area was merged in yellow (c). Upon heat shock for 10 min or 30 min, both G3BP (d, g) and Tudor‐SN (e, h) were visualized in the same cytoplasmic foci, and the merged picture showed the co‐localization of the two proteins (f, i). At 60 min, the SGs were characterized as large granule aggregates containing both Tudor‐SN (k) and G3BP (j), which formed around the nucleus (l). To confirm that the cytoplasmic foci are stress granules, we also detected the distribution of Tudor‐SN with another marker protein of stress granule, TIA‐1. As shown in Fig. 3B, after heat shock, both Tudor‐SN (green, d) and TIA‐1 (red, e) were found in the same cytoplasmic foci (yellow, f). These data verified that the Tudor‐SN protein is bona fide novel member of stress granules.

Furthermore, we overexpressed GFP‐G3BP and RFP‐Tudor‐SN in HeLa cells, and then performed kinetic experiments to monitor the assembly of SGs in living cells treated with 0.5 mM sodium arsenite. The results in Fig. 4 showed the co‐ordinated recruitment of G3BP (green) and Tudor‐SN (red) to SGs during the assembly process. Without stimulation, GFP‐G3BP (green, a) and RFP‐Tudor‐SN (red, b) were distributed and co‐localized (merged yellow, c) mainly in the cytoplasm. As reported earlier [3], we also observed that overexpression of G3BP efficiently trigger the assembly of SGs even in the absence of stress stimuli (Fig. 4a, white arrows in the inset), and ectopically expressed Tudor‐SN could also recruit into the same foci (Fig. 4b, white arrows in the inset). With arsenite treatment for 5 min, G3BP (green, d) and Tudor‐SN (red, e) were gradually re‐distributed into some small stress granules which merged into yellow foci (f). At 10 min, lots of stress granules were formed which contained both Tudor‐SN (red, h) and G3BP (green, g). After 20 min, the small SGs fused into larger ones which showed double positive staining of G3BP (green, g, j) and Tudor‐SN (red, h, k) as indicated in the enlarged areas. In summary, in response to stress stimuli, the stress granules containing both Tudor‐SN and G3BP are aggregated first in small size dispersed in the cytoplasm, and then in large granules around the nucleus. These observations demonstrate that Tudor‐SN and G3BP are recruited simultaneously to the SGs in response to stress conditions.

Next we investigated whether the SN domain of Tudor‐SN participates in the SGs assembly. HeLa cells were transfected with GFP‐G3BP and RFP‐Tudor‐SN, RFP‐SN, or RFP‐TSN, respectively. After 24 h, the cells were seeded on glass cover slips, and then treated with either sodium arsenite or heat shock. The images were collected with confocal microscope. In normal cells, RFP‐Tudor‐SN (Fig. 5 A, b) and RFP‐SN were primarily distributed in the cytoplasm (Fig. 5B, b), RFP‐TSN was found to mainly localize in the nucleus (Fig. 5C, b). Under two different stress stimuli, both RFP‐Tudor‐SN (Fig. 5A, e and h) and RFP‐SN (Fig. 5B, e and h) were efficiently recruited into SGs with GFP‐G3BP (merged yellow, f, i). However, as shown in Fig. 5C, RFP‐TSN domain did not redistribute to the cytoplasmic foci (e, h) with GFP‐G3BP (d, g). Fig. 5C, (f and i) demonstrated the separate localization of RFP‐TSN and GFP‐G3BP. This data is consistent with the previous results showing the interaction between G3BP and SN domain, and the lack of interaction with TSN domain. These results demonstrate that the SN domain of Tudor‐SN protein is involved in the formation of SGs. To exclude the possibility that the co‐localization was caused by GFP and RFP, HeLa cells transfected with empty vector of pEGFP‐C1 and p Cherry‐C1 were also treated with heat shock or arsenite. As shown in Fig. 5D, the localization of RFP or GFP alone was not affected by the stress treatment (c, f, i), and the insets showed that no cross‐detection occurred between the green and red channels.

To investigate the significance of Tudor‐SN in the formation of stress granules, we performed knockdown experiments with siRNAs which directed against Tudor‐SN or scrambled siRNA as control. As shown in Fig. 6 A, transfection of Tudor‐SN siRNAs significantly reduced the expression of endogenous Tudor‐SN protein by about 80% (upper panel) comparing with the scrambled siRNA control, but has no effect on the abundance of G3BP (middle panel) (Fig. 6D) or GAPDH (lower panel). And the knockdown of Tudor‐SN protein inhibited the cell proliferation (Fig. 6B), but did not affect the cell viability. The transfected HeLa cells were seeded on glass cover slips, and incubated at 45 °C for 60 min or treated with 0.5 mM sodium arsenite for 45 min. Immunofluorescence experiments were performed using monoclonal anti‐G3BP antibody and polyclonal anti‐Tudor‐SN antibody. As shown in Fig. 6C, in control cells, both G3BP (a) and Tudor‐SN (b) were visualized in the cytoplasm of all the cells. Upon heat shock or arsenite treatment, G3BP and Tudor‐SN were efficiently recruited into the large SGs around the nucleus in all the HeLa cells (f and i). With knockdown of endogenous Tudor‐SN protein, the green staining of G3BP protein was clearly observed in all the cells, but the red staining of Tudor‐SN protein was not observed in about 70% percent of the cells. Photographs of three representative cells were selected for illustration in Fig. 6D. In the HeLa cells with knockdown of endogenous Tudor‐SN protein (e, h), the heat shock or arsenite treatment caused the formation of small SGs (G3BP in green, d, g) dispersed throughout the cytoplasm, which cannot aggregate into large foci as in the control cells (Fig. 6C, d and g). These results indicate that although Tudor‐SN may not be an essential factor to trigger the formation of SGs, it is likely to play important roles in the aggregation of the SGs.