Endogenous TDP-43, but not FUS, contributes to stress granule assembly via G3BP

We have previously demonstrated that TDP-43 depletion leads to a defect in the timing of SG formation in response to acute oxidative stress induced by sodium arsenite (SA) treatment [12]. In addition, we reported that SG in TDP-43 depleted cells assayed immediately after the stress were 43% smaller than control cells [12], but the number of SG per cell is comparable to siControl cells (73.4 ± 3.6 vs. 82.2 ± 6.4, p = 0.15). These data suggest a potential role for TDP-43 in SG dynamics. It is well documented that following initial SG formation, SG slowly reduce in number but each individual SG increases in size before resolving. This SG size increase is attributed to the fusion of multiple TIA-1 oligomers and is referred to as SG assembly [8]. However, the mechanism(s) which mediate this step is unclear.

Building on our previous published data in which we validated two independent TDP-43 specific siRNAs, we selected one of these siRNAs for transfection into HeLa cells and then subsequently stressed the cells for 30 min with 0.5 mM SA followed by recovery with fresh media. Cells were subsequently fixed at 60 and 90 min and SG were visualized with TIA-1 labelling (Figure 1A). In HeLa cells transfected with control siRNA (siControl), we observed an average of 84 ± 5 SG per cell with an average size of 1.9 ± 0.1 μm² at 60 min, followed by a reduction in SG number (60 ± 6 SG, p = 0.005) and an increase in SG size (to 3.0 ± 0.1 μm², p = 0.009) at 90 min (Figure 1B). This simultaneous decrease in SG number and increased SG size is due to the secondary aggregation of TIA-1 oligomers and is referred to as SG assembly [8]. In cells depleted of endogenous TDP-43, SG assembly is blocked such that SG number and size remains unchanged at both time points examined (Figure 1B). A second TDP-43 targeted siRNA also yielded the same results (Additional file 1: Figure S1). Moreover, this blockade in SG assembly was also true in SK-N-SH cells, a neuronal-like cell line (Figure 1C).

To further evaluate whether TIA-1 oligomerization was disrupted in SG assembly, we performed a protease sensitivity assay. Briefly, SG assembly is reflected as increased TIA-1 aggregation and thus increased protease resistance [9, 36]. Cells transfected with TDP-43 or control siRNA were stressed with SA and then permitted to recover for 60 min before cytoplasmic fractions were subject to a titration of Proteinase K (PK) and blotted for TIA-1. In siControl cells, we observed an increased resistance of TIA-1 to proteolysis between stressed and unstressed cells (Figure 1D, compare lanes 5 and 11; Figure 1E). In contrast, TIA-1 protease-sensitivity in SA-treated TDP-43 depleted cells was comparable to untreated cells, indicating decreased TIA-1 aggregation and thus impaired SG assembly (Figure 1D, compare lanes 11 and 23, red arrows; Figure 1E). Collectively, these experiments demonstrate that endogenous TDP-43 is required for normal SG assembly.

We have previously reported that G3BP is down-regulated in cells depleted of TDP-43 via two independent siRNAs [12]. We hypothesized that the SG assembly defect observed in TDP-43 depleted cells could be attributed to decreased G3BP expression. To address this, we reduced G3BP levels using a gene-specific siRNA to levels comparable to cells treated with TDP-43 siRNA (remaining G3BP protein in siG3BP vs. siTDP-43 in HeLa cells: p = 0.08; in SK-N-SH: p = 0.32; Figure 2A, B). Importantly, we noted that G3BP depletion does not impact TDP-43 expression (HeLa: p = 0.071, SK-N-SH: p = 0.21; Figure 2A, B). G3BP is considered essential for SG initiation based on studies where overexpression of G3BP is sufficient to nucleate SG [10]. However, unlike TIA-1, which is required for SG initiation, the impact of G3BP on SG initiation by either siRNA or genetic deletion has not been reported. To define if G3BP is necessary for SG initiation at endogenous expression levels, we counted the number of SG per cell and measured SG size in G3BP-depleted conditions immediately following a 30 min SA exposure. We observed that the number of SG per cell was equivalent in siControl and siG3BP cells (113 ± 3 vs. 117 ± 11 SG per cell, p = 0.373, Additional file 2: Figure S2). Moreover, we did not detect a difference in initial SG size (1.7 ± 0.1 vs. 1.8 ± 0.1 μm², p = 0.398, Additional file 2: Figure S2). Since SG composition can vary between cell types, we also evaluated SK-N-SH cells and observed strikingly similar results with respect to SG number (55 ± 8 vs. 63 ± 13 SG per cell, p = 0.332) and size (1.43 ± 0.01 vs. 1.42 ± 0.03 μm², p = 0.366). Together, these data indicate G3BP is a non-essential factor in the initiation of SG in response to acute oxidative stress induced by SA.

To evaluate the role of G3BP in SG assembly, we quantified the number of SG per cell as well as individual SG size at 60 and 90 min after SA treatment. Visual inspection of SG by confocal suggested a difference in SG number and size in siG3BP cells compared to controls. Quantification of individual SG in these cells revealed that the number and the size of SG in siG3BP HeLa cells is comparable to siControl cells at 60 min (p = 0.243 and p = 0.400, respectively; Figure 2C). In contrast, while average SG size increases 1.6-fold in siControl cells (p = 0.004) as expected, siG3BP cells remain constant at 1.9 ± 0.1 μm². We obtained similar results with a second siRNA for G3BP (Additional file 1: Figure S1) and in SK-N-SH cells (Figure 2D) indicating that the requirement of G3BP for SG assembly is not a cell-type specific event. Taken together, our data demonstrate G3BP as an important factor in SG assembly, but not SG initiation as previously suggested.

Our previous work indicated that TDP-43 depletion using two independent siRNAs resulted in a robust and reproducible decrease in G3BP expression [12]. To investigate whether the defect in SG assembly observed in TDP-43 depleted cells could be attributed to G3BP, we transiently expressed GFP-tagged G3BP in TDP-43 depleted cells, at a level comparable to endogenous G3BP expression (Figure 3A). Using our conditions, where we transfected 0.3 μg of G3BP-GFP cDNA for 24 hr, we detected cells of both high and low expression by immunofluorescence. Consistent with a previous report, G3BP overexpression at high levels induces the spontaneous formation of SG [10] (Figure 3B, asterisks) such that 60 ± 0.4% of transfected cells formed SG in the absence of SA treatment (Figure 3C). Interestingly, we noted that the depletion of TDP-43 resulted in a 1.9-fold reduction in the number of cells with spontaneous SG formation suggesting that endogenous TDP-43 participates in this process (p = 0.0006).

In order to address our initial hypothesis that G3BP mediates SG assembly in TDP-43 depleted cells, we selected transfectants with low G3BP-GFP expression for further analysis. Low expression was defined via measurement of the GFP signal in individual cells. Cells having an average pixel intensity below 60 arbitrary fluorescent units were uniquely included in the subsequent quantification. This population of low-expressing G3BP-GFP cells do not exhibit spontaneous SG formation (Figure 3B, arrowheads). As expected, cells co- transfected with control siRNA and control GFP cDNA undergo normal SG assembly while siTDP-43 cells transfected with GFP cDNA demonstrate a defect in SG assembly (Figure 3D-F). The low expression of G3BP-GFP in TDP-43 depleted cells restored the number and size of SG at 90 min to levels equivalent to controls, indicating a restoration of SG assembly (Figure 3E, F). Thus, the defect in SG assembly in TDP-43 depleted cells can be rescued by G3BP expression. These data indicate that G3BP functions downstream of TDP-43 in SG assembly.

There have been several reports that a second ALS-causing gene FUS/TLS, which is an hnRNP like TDP-43, localizes to SG in response to oxidative stress [26‐28]. All of these studies demonstrated SG localization of over-expressed FUS protein, but do not evaluate the contribution of endogenous FUS to this process. Since there is suggestion that TDP-43 and FUS participate in similar biological processes [37‐40], we investigated the potential role of endogenous FUS in SG dynamics in response to oxidative stress using siRNA depletion in HeLa cells. We did not observe any difference in the number of cells able to form TIA-1 labelled SG in response to SA treatment at any time point examined (data not shown). Furthermore, SG initiation (number and size at 30 min) was undisturbed in cells depleted of FUS (Additional file 2: Figure S2). Similarly, SG assembly proceeded normally as SG number and size were similar between siControl and siFUS cells at 60 and 90 min post-stress (Figure 4A, B). These data were confirmed with a second siRNA specific for FUS (Additional file 1: Figure S1). Since TDP-43 depletion can alter the expression of TIA-1 and G3BP [12], we also evaluated the levels of these proteins in siFUS cells. FUS depletion (92% in HeLa and 68% in SK-N-SH) did not impact TIA-1 or G3BP protein expression in HeLa or SK-N-SH cells (Figure 4C, D). We also noted that FUS depletion does not impact the expression of endogenous TDP-43 (Figure 4C, D) nor does it impact TDP-43 localization to SG (data not shown). We conclude that endogenous wild type FUS does not participate in SG initiation or assembly in response to acute oxidative stress.

SG comprise one arm of an integrated cellular stress response. Previous data indicate that a failure in SG dynamics (meaning formation, assembly, or disassembly) contributes to poor cellular recovery to acute stress [13‐15]. We have previously shown that TDP-43 contributes to cell survival 24 hr after stress exposure [12]. To determine if this was a short- or long-term requirement and whether this was exclusive to TDP-43, we extended these studies such that cells transfected with specific siRNAs for 48 hr were stressed (SA, 30 min) and then assessed for cell viability at 24 and 48 hr by trypan blue exclusion. In all cases, robust depletion of TDP-43, G3BP, and FUS was achieved by the respective siRNAs (Figure 5A, C). In the case of TDP-43 depleted HeLa cells, we noted a 1.9-fold increase (p = 0.01) in cell death in unstressed conditions at 48 hr while no significant differences in viability were detected in siG3BP or siFUS cells (Figure 5B). When challenged with SA treatment, siTDP-43 cells fared worse than siControl cells 24 hr later and cell death was further increased to 2.3-times siControl cells at 48 hr (24 hr: p = 0.02; 48 hr: p = 0.01, Figure 5B). Depletion of FUS had no impact on cell viability at any time point, consistent with our data that endogenous FUS is not required for a SG response. Surprisingly, depletion of G3BP did not impact cellular recovery to oxidative stress in HeLa cells (Figure 5B). However, additional experiments in the neuronal-like cell line SK-N-SH revealed similarly increased cell vulnerability 48 hr post-SA treatment in both TDP-43 and G3BP depleted cells (siTDP-43: 2.3-fold, p = 0.019; siG3BP: 2.6-fold, p = 0.005) (Figure 5D). In contrast, as in HeLa cells, FUS depletion had no effect on survival in these neuronal-like cells. Thus, neuronal-like cells seem to be more vulnerable to a defect in SG assembly when either TDP-43 or G3BP is depleted. In both lines, cells transfected with control siRNA exhibited a background level of cell death that was constant throughout the experiments.

TDP-43 depleted cells have not only a down-regulation of G3BP but also a 30% up-regulation of TIA-1 protein levels (and 2.6-fold increase in TIA-1 transcripts) [12] that is not observed in G3BP depleted cells (Figure 2A). Since G3BP siRNA failed to recapitulate the same level of toxicity as TDP-43 depletion in HeLa cells, we wondered if there was a differential sensitivity to elevated TIA-1 expression between cell types. To address this, we transiently transfected cells with a low concentration of GFP-tagged TIA-1 for 24 hr (Figure 6A, B). At 24 hr post-transfection, we observed that cells with high GFP-TIA-1 expression formed puncta resembling SG in the absence of exogenous stress, in agreement with a previous report [9]. (However, TIA-1 aggregates were not detected at 48 hr, data not shown.) Cells were then briefly stressed with SA (30 min) and the media was replaced so as to permit cellular recovery for 24 hr. In order to avoid inducing non-specific cell death attributed to supraexpression of TIA-1, we used a very low concentration of cDNA in these experiments. This resulted in a low transfection efficiency. Thus, in order to selectively assess cell death in GFP⁺ transfectants (and thus exclude non-transfectants), we employed a flow cytometry method to assess cell death via Annexin V labelling specifically in GFP-expressing cells. By this method, we observed that GFP-TIA-1 transfected HeLa cells have a 2.3-fold increased toxicity following SA treatment (p = 0.01, Figure 6C). Also, we note that in unstressed/basal conditions, GFP-TIA-1 transfected cells have nearly twice the toxicity as control GFP transfected cells (p = 0.005, Figure 6C). Note, while TIA-1 has been previously described as cytotoxic, the toxicity observed here is independent of the level of TIA-1 overexpression as demonstrated by independently gating populations of high and low GFP-TIA-1 expression (Additional file 3: Figure S3). In contrast, control GFP transfected cells were comparable to untransfected cells in both basal and stressed conditions. When we performed the same experiment in SK-N-SH cells, we noted that TIA-1 over-expression did not sensitize cells to SA (Figure 6C). Note, the apparent difference in TIA-1 expression detected by immunoblot likely arises from a reduced transfection efficiency of SK-N-SH cells (compare images in Figure 6A), while FACS analysis was restricted to cells expressing either GFP or GFP-tagged fusion protein. Taken together, these data suggest that low-level overexpression of TIA-1 can be toxic to certain cell types and this toxicity can be exacerbated by the addition of an exogenous stress.

HeLa cells were grown in Dulbecco’s modified Eagle medium (Invitrogen, Carlsbad, CA, USA), supplemented with 10% fetal bovine serum and 1% glutamine. For transfection of siRNA, 125 pmol of custom siRNAs were transfected with Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s instructions. siRNA sequences used were: human FUS/TLS: 5^′-GGCCAAGAUCAAUCCUCCAUGAGUA-3^′, and #2 5^′-CGGGACAGCCCAUGAUUAAUUUGUA-3^′; human G3BP1: 5^′-ACAUUUAGAGGAGCCUGUUGCUGAA-3^′, and #2 5^′-GCGCAUUAACAGUGGUGGGAAAUUA-3^′; human TDP-43: 5^′-AAGCAAAGCCAAGAUGAGCCUUUGA-3^′, and #2 5^′- AAGAUGAGAACGAUGAGCCCAUUGA-3^′, and negative control low GC siRNA (catalogue number 12935-200, Invitrogen, Carlsbad, CA, USA). Cells were transfected at 30–50% confluency and collected after 72 hr. For transfection of cDNA, 0.3 μg GFP tagged TIA-1 (provided by J. Goodier, Johns Hopkins University, MD) or 0.3 μg of G3BP-GFP (provided by J. Tazi, Montpellier, France) were transfected with Lipofectamine LTX (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. Cells were transfected with cDNA 48 hr after siRNA treatment. Cells were collected after 24 hr of treatment. Cells were treated with 0.5 mM sodium arsenite (30 min, 37°C; Sigma). Recovery was initiated by changing the media. Cells were fixed at the indicated time points.

Cells grown on cover slips were fixed in 1% formaldehyde in phosphate buffered saline (PBS) and subsequently permeabilized with 0.1% Triton X-100 in PBS for 15 min. Cover slips were blocked with 0.1% bovine serum albumin (BSA) in PBS for 15 min and incubated with antibodies to TDP-43 (1:400; Proteintech, Chicago, IL, USA), TIA-1 (1:100; Santa Cruz, Santa Cruz, CA, USA), G3BP (1:400; BD Biosciences, Franklin Lakes, NJ, USA), diluted in blocking buffer for 1 hr at room temperature. Cover slips were subsequently washed once with 0.1% Triton X-100 in PBS and twice with 0.1% BSA in PBS. Labelling was visualized with the fluorescently conjugated secondary antibodies donkey anti-mouse Texas Red (1:200; Jackson Immunochemicals, West Grove, PA, USA), donkey anti-rabbit FITC (1:200; Jackson Immunochemicals, West Grove, PA, USA), and donkey anti-goat Alexa 633 (1:800; Invitrogen, Carlsbad, CA, USA). For some experiments, cells were stained with TO-PRO 3 (1:300; Invitrogen, Carlsbad, CA, USA) 10 min at room temperature. Cover slips were washed tree times in PBS and mounted with ProLong Antifade reagent (Invitrogen, Carlsbad, CA, USA). Images were collected on a Leica SP5 confocal microscope.

ImageJ was used for quantification of SG parameters. SGs were identified by TIA-1 staining and cells were scored as positive when they had at least two foci of a minimal size of 0.75 μm². Automatic recognition was done by Image J using the following parameters, SGs ranging from 0.75 to 5 μm², randomly selected in 10 cells per condition. The average SG size of at least 100 SG is presented. For G3BP-GFP experiments, cells with an average pixel intensity of 60 arbitrary fluorescent units were selected for analysis. Control cells at this level of G3BP-GFP expression did not exhibit SG formation in basal conditions.

Cells were collected in ice-cold PBS, lysed in 10 mM Tris pH 8.0, 50 mM NaCl, 0.5 M sucrose, 0.1 mM EDTA, 0.5% Triton X-100, 1 mM DTT. Cytoplasmic fractions were recovered as supernatants after centrifugation at 230 g for 10 min and quantified with the BCA Protein Assay Kit (Thermo Scientific, Waltham, MA, USA). 50 μg of cytoplasmic extract were incubated with increasing concentrations of Proteinase K (0 - 1.6 mg/ml) for 5 min at 37°C. Proteinase K was neutralised by 10 mM PMSF for 10 min on ice [36]. Samples were then resuspended in 1X Laemmli sample buffer.

Cell viability in siRNA depleted cells was scored via trypan blue exclusion at the indicated times after stress. Cell viability/toxicity in GFP-TIA-1 transfected cells was measured using the Annexin V/7-AAD labelling kit (BD Biosciences, Franklin Lakes, NJ, USA) according to the manufacturer’s instructions In these experiments, cells were gated for GFP and Annexin V⁺ cells are reported in the figure. Data acquisition was performed on a BD LSR II flow cytometry (BD Biosciences) and data was analyzed with FlowJo (Treestar, Ashland, OR).

In all cases, data was compared via a two-tailed Student t-test using Microsoft Excel.