TDP-43 is recruited to stress granules in conditions of oxidative insult

Authors

  • Claudia Colombrita,

    1. Department of Neurology and Laboratory of Neuroscience, ‘Dino Ferrari’ Center, Università degli Studi di Milano – IRCCS Istituto Auxologico Italiano, Milan, Italy
    Search for more papers by this author
  • Eleonora Zennaro,

    1. Department of Neurology and Laboratory of Neuroscience, ‘Dino Ferrari’ Center, Università degli Studi di Milano – IRCCS Istituto Auxologico Italiano, Milan, Italy
    Search for more papers by this author
  • Claudia Fallini,

    1. Department of Neurology and Laboratory of Neuroscience, ‘Dino Ferrari’ Center, Università degli Studi di Milano – IRCCS Istituto Auxologico Italiano, Milan, Italy
    2. Department of Cell Biology, Emory University School of Medicine, Atlanta, Georgia, USA
    Search for more papers by this author
  • Markus Weber,

    1. Neuromuscular Diseases Unit/ALS Clinic, Kantonsspital St. Gallen, St.Gallen, Switzerland
    Search for more papers by this author
  • Andreas Sommacal,

    1. Institute of Pathology, Kantonsspital St. Gallen, St.Gallen, Switzerland
    Search for more papers by this author
  • Emanuele Buratti,

    1. International Centre for Genetic Engineering and Biotechnology (ICGEB), AREA Science Park, Trieste, Italy
    Search for more papers by this author
  • Vincenzo Silani,

    1. Department of Neurology and Laboratory of Neuroscience, ‘Dino Ferrari’ Center, Università degli Studi di Milano – IRCCS Istituto Auxologico Italiano, Milan, Italy
    Search for more papers by this author
    • 1

      Vincenzo Silani and Antonia Ratti are Joint Senior Authors.

  • Antonia Ratti

    1. Department of Neurology and Laboratory of Neuroscience, ‘Dino Ferrari’ Center, Università degli Studi di Milano – IRCCS Istituto Auxologico Italiano, Milan, Italy
    Search for more papers by this author
    • 1

      Vincenzo Silani and Antonia Ratti are Joint Senior Authors.


Address correspondence and reprint requests to Antonia Ratti, Department of Neurology and Laboratory of Neuroscience, IRCCS Istituto Auxologico Italiano, Via Zucchi, 18 – 20095 Cusano Milanino, Milan, Italy. E-mail: antonia.ratti@unimi.it

Abstract

Transactive response DNA-binding protein 43 (TDP-43) forms abnormal ubiquitinated and phosphorylated inclusions in brain tissues from patients with amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration. TDP-43 is a DNA/RNA-binding protein involved in RNA processing, such as transcription, pre-mRNA splicing, mRNA stabilization and transport to dendrites. We found that in response to oxidative stress and to environmental insults of different types TDP-43 is capable to assemble into stress granules (SGs), ribonucleoprotein complexes where protein synthesis is temporarily arrested. We demonstrated that a specific aminoacidic interval (216–315) in the C-terminal region and the RNA-recognition motif 1 domain are both implicated in TDP-43 participation in SGs as their deletion prevented the recruitment of TDP-43 into SGs. Our data show that TDP-43 is a specific component of SGs and not of processing bodies, although we proved that TDP-43 is not necessary for SG formation, and its gene silencing does not impair cell survival during stress. The analysis of spinal cord tissue from ALS patients showed that SG markers are not entrapped in TDP-43 pathological inclusions. Although SGs were not evident in ALS brains, we speculate that an altered control of mRNA translation in stressful conditions may trigger motor neuron degeneration at early stages of the disease.

Abbreviations used:
ALS

amyotrophic lateral sclerosis

FTLD

frontotemporal lobar degeneration

FUS/TLS

Fusion/Translocated in LipoSarcoma

HuR

Hu R antigen

P-bodies

processing bodies

RBP

RNA-binding protein

RNP

ribonucleoprotein

RRM

RNA-recognition motif

SG

stress granule

TDP-43

Transactive response DNA-binding protein 43

TIA-1

T cell-induced antigen 1

TIAR

TIA-related

Transactive response DNA-binding protein 43 (TDP-43) is an ubiquitously expressed RNA-binding protein (RBP) belonging to the heterogeneous ribonucleoprotein (RNP) family and containing two RNA-recognition motif (RRM) domains for target RNA binding and a glycine-rich C-terminal tail for protein-protein interaction (Ayala et al. 2005). The biological role of TDP-43 is associated both to transcriptional regulation and to post-transcriptional control of RNA processing, ranging from splicing to mRNA stabilization and transport (Buratti and Baralle 2008).

TDP-43 has recently emerged as the major neuropathological hallmark of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) (Arai et al. 2006; Neumann et al. 2006). Although distinct neuronal populations and brain areas selectively degenerate in ALS and in FTLD, these two disorders represent a clinical continuum as FTLD patients may also develop motor neuron disease and cognitive deficits have been observed in ALS. However, TDP-43 inclusions have been identified also in Alzheimer’s, Parkinson’s and Huntington’s diseases, suggesting it may be a common marker for many neurodegenerative disorders (Amador-Ortiz et al. 2007; Nakashima-Yasuda et al. 2007; Schwab et al. 2008). The fact that TDP-43 protein may be primarily and specifically involved in ALS onset has recently emerged from genetic studies showing TARDBP gene mutations in familial and sporadic ALS cases (Sreedharan et al. 2008; Corrado et al. 2009). Whether TDP-43 protein aggregation is a pathogenic event that triggers neuronal degeneration or whether TDP-43-positive inclusions are the consequence of a neuroprotective mechanism still remains an issue to be addressed.

In condition of oxidative stress the regulation of gene expression at post-transcriptional level, mediated by RBPs, is known to be impaired. In particular, when a sub-lethal oxidative stress is induced in vitro, there is an immediate block of the translation machinery with sequestration of the actively-translating mRNAs and distinct RBPs to cytoplasmic foci, called stress granules (SGs) (Anderson and Kedersha 2008). SGs represent a protective mechanism to bypass the cellular insult as the majority of mRNAs is silenced in these macromolecular structures in stalled 48S ribosomal complexes, while only specific and essential transcripts (i.e. Hsp70) are maintained in active translation (Anderson and Kedersha 2002; Kedersha and Anderson 2002). During stress, SGs are in dynamic equilibrium between polysomes and processing bodies (P-bodies), the latter being constitutive RNP complexes where both mRNA degradation and microRNA-mediated translational arrest take place (Kedersha et al. 2005, 2008). It is experimentally proven that once the insult is removed, these RNP complexes soon disaggregate in favour of a parallel polysome re-assembly and mRNA translation re-initiation. The molecular mechanisms and the signalling pathways triggering SG formation have been characterized, as well as the nature of distinct cellular insults which can induce these structures, including oxidative stress, proteasome inhibition, osmotic and heat shocks (Kedersha and Anderson 2007).

In addition to mRNA, several RBPs have been described so far to be components of these cytoplasmic foci in condition of cellular insult. T cell-induced antigen 1 (TIA-1) and TIAR (TIA-related) are the essential RBPs which promote SG assembly because of their prion-like domains that favour aggregation of other RBPs/proteins in granules together with their target mRNAs. SGs also contain PolyA-Binding Protein 1 (PABP-1), the Embryonic Lethal Abnormal Vision (ELAV) family member Hu R antigen (HuR), Survival Motor Neuron (SMN), Ras-GAP SH3 domain-binding protein (G3BP), Staufen, and Fragile X Mental Retardation Protein (FMRP) RBPs, together with the ribosomal 48S pre-initiation complex, early translation initiation factors, microRNA-associated Argonaute proteins, p54/Rck helicase, XRN1 exonuclease and cytoskeletal proteins (Anderson and Kedersha 2008).

The aim of our study was to investigate whether the RBP TDP-43 participates to the assembly of SGs in condition of cellular stress in a motoneuronal cell line and whether such cytoplasmic RNP complexes are also present in the spinal cord and/or in the TDP-43-positive pathological inclusions of ALS patients.

Experimental procedures

Cell culture and treatments

The motoneuronal cell line NSC34 (a kind gift of N.R. Cashman, University of British Columbia, Vancouver, Canada) was cultured as previously reported (Ratti et al. 2008). NSC34 cells were exposed to 0.5 mM sodium arsenite for 30 min or pre-treated with emetine (20 μg/mL for 2 h) or puromycin (20 μg/mL for 4 h) as described (Kedersha and Anderson 2007). MG132 (10 μM) was used for 4 h as reported (Mazroui et al. 2007). All reagents were purchased from Sigma (Milan, Italy). For heat shock experiments, cells were incubated at 44°C for 30 min overlaid with mineral oil.

Immunocytochemistry

Cells were fixed with 4% paraformaldehyde in phosphate buffered saline for 15 min, permeabilized with cold methanol and 0.2% Triton X-100, and blocked with 10% normal goat serum solution (Vector Laboratories, Burlingame, CA, USA). Incubation with primary antibodies (TDP-43, 1 : 500, ProteinTech Group, Manchester, UK; TIAR, 1 : 100, BD Transduction Laboratories, Milan, Italy; HuR, 1 : 500, Molecular Probes, Milan, Italy; Nova1, 1 : 200, Upstate Biotechnology, Milan, Italy; FLAG, 1 : 1200, Sigma) was performed in blocking solution for 1 h at 37°C. The fluorescent-tagged secondary antibodies Alexa Fluor 488 and 555 (1 : 500, Invitrogen, Milan, Italy) were used for detection and nuclei were visualized by 4′-6-diamidino-2-phenylindole (DAPI) staining (Roche, Milan, Italy). As a negative control, primary antibodies were replaced by normal goat serum. Slides were mounted with Fluorsave (Calbiochem, La Jolla, CA, USA) and acquired with a widefield microscope (DMIRE2/HCS, Leica Microsystems, Wetzlar, Germany).

Protein extraction, immunoprecipitation and western blotting

NSC34 cells were homogenized in lysis buffer (150 mM NaCl, 20 mM Tris-HCl pH7.4, 1% Triton X-100, protease inhibitor cocktail), centrifuged at 12 000 g for 15 min at 4°C and supernatants were collected. Proteins from cytoplasmic and nuclear fractions were obtained with the ProteoJETTM kit (Fermentas, Milan, Italy) following the manufacturer’s instructions. For immunoprecipitation experiments, 30 μL protein G Sepharose-beads pre-coated for 6 h with 1.5 μg of the selected antibody were incubated with 300 μg cytoplasmic protein lysate in NT2 buffer [50 mM Tris-HCl pH7.4, 15 mM NaCl, 1 mM MgCl2, 0.05% NP-40 (Sigma)], containing 400 U RNase inhibitor, 1 mM dithiothreitol and 20 mM EDTA. After overnight incubation and four washes in NT2 buffer, recovered proteins were resolved on 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred to nitrocellulose membrane. Western blot and immunoprecipitation assays were performed with HuR, TIAR, TIA-1 and α-tubulin (all from Santa Cruz Biotechnology, Santa Cruz, CA, USA), TDP-43 and p84 (Abcam, Cambridge, UK) antibodies.

Plasmid constructs, siRNA and transfections

NSC34 cells were cultured on glass cover-slips and transiently transfected with Lipofectamine 2000 (Invitrogen) following the manufacturer’s instructions. After 48-h transfection cells were exposed to 0.5 mM arsenite for 30 min prior to be processed for immunocytochemistry. The FLAG-tagged full-length and deleted (ΔRRM1, ΔC, 1–315) TDP-43 plasmids used for transfection were previously described (Ayala et al. 2008). The Dcp1-enhanced Green Fluorescent Protein (EGFP) and TIA-1-enhanced Yellow Fluorescent Protein (EYFP) constructs were kindly provided by Dr. P. Macchi (University of Trento, Italy). For gene silencing experiments the following siRNA duplexes were used: 5′-GCAAAGCCCAGACGAGCCUdTdT-3′ for mouse TDP-43; siGenome Non-Targeting siRNA #2 (Dharmacon, Lafayette, CO, USA) against the firefly luciferase gene as a non-specific control.

Immunohistochemistry

Paraffin-embedded 8 μm-thick sections from cervical spinal cord of three sporadic ALS patients were processed and re-hydrated following a standard protocol. After pre-treatment with 10 mM sodium citrate buffer pH6 for 20 min at 80°C, tissue sections were permeabilized with 0.3% Triton X-100 and blocked with 10% normal goat serum for 20 min. Double staining with TDP-43 (1 : 200) and TIAR (1 : 60, BD Transduction Laboratories) or HuR (1 : 50, Santa Cruz Biotechnology) antibodies was performed overnight at 4°C in phosphate buffered saline and Alexa Fluor-conjugated secondary antibodies (1 : 500) were used for detection. Nuclei were visualized by DAPI staining. As a negative control, primary antibodies were replaced by normal goat serum. Slides were acquired with a confocal microscope (LSM510 META, Zeiss, Jena, Germany).

Results

TDP-43 forms stress granules after arsenite treatment

We investigated whether TDP-43, like other RBPs, is able to assemble into SGs in condition of oxidative stress by treating the immortalized motoneuron-like NSC34 cells with 0.5 mM arsenite for 30 min as previously described (Kedersha and Anderson 2007). As SG markers in immunofluorescence assays we used antibodies against TIAR, which in physiological conditions shows both a nuclear and cytosolic distribution (Fig. 1a), and HuR which is predominantly nuclear, like TDP-43 (Fig. 1b). After arsenite treatment we observed the formation of cytoplasmic foci which stained positive for TIAR (80% of cells) and HuR (60% of cells). A sub-group of cells with TIAR or HuR-positive SGs (30% and 50%, respectively) showed co-localization also with TDP-43 protein in cytoplasmic granules (Fig. 1a and b). To further demonstrate that TDP-43-positive foci were SGs, we incubated NSC34 cells with arsenite in the presence of two distinct pharmacological inhibitors of translation, emetine and puromycin, known to prevent or favour SG formation, respectively (Kedersha and Anderson 2007). Induction of oxidative stress in the presence of emetine was not able to trigger SG assembly (Fig. 1c), whilst with puromycin pre-conditioning we observed the formation of SGs which stained positive for HuR and TDP-43 proteins (Fig. 1d). Cell treatment with a milder dose of arsenite for a prolonged period of time (15 μM for 20 h) did not lead to the formation of SGs, but to a diffuse redistribution of TDP-43 protein in the cytoplasm (data not shown).

Figure 1.

 Recruitment of TDP-43 protein into SGs after arsenite treatment in vitro. (a) Immunofluorescence images of TIAR (green) and TDP-43 (red) proteins in untreated (Untr) and arsenite-treated (Ars) NSC34 cells. Merged image shows co-localization signals in cytoplasmic SGs after induction of oxidative stress. (b) Immunocytochemistry of HuR (green) and TDP-43 (red) proteins in untreated and Ars-treated cells. The concurrent use of arsenite and the translation inhibitors emetine (c) and puromycin (d) prevented or favoured SG formation, respectively. Nuclei were counter-stained by DAPI (blue). Scale bar, 10 μm.

TDP-43 distributes into the cytoplasm following oxidative stress

We confirmed the immunofluorescence data by biochemical analyses of NSC34 protein extracts after arsenite insult. Western blot assays revealed no change in the total amount of TDP-43 protein after induction of oxidative stress, as well as of the other two SG markers HuR and TIAR (Fig. 2a). What we observed was a redistribution of the TDP-43 protein from a main nuclear localization to a cytoplasmic one (Fig. 2b). TIA-1 and TIAR-containing RNP complexes were recovered by specific immunoprecipitation from the cytosolic fraction and probed for the presence of TDP-43 and HuR RBPs. Arsenite treatment significantly increased the association of the two proteins with both TIA-1 and TIAR SG markers in the cytoplasm of NSC34 cells (Fig. 2c).

Figure 2.

 Biochemical analyses of NSC34 protein extracts after arsenite insult. (a) Representative western blot showing the total content of TDP-43, HuR and TIAR proteins in NSC34 control cells (Untr) and after 30-min treatment with arsenite (Ars). α-tubulin was used for sample normalization. (b) Immunoblot of TDP-43 protein in nuclear and cytoplasmic fractions in control (Untr) and Ars-treated NSC34 cells. p84 and α-tubulin were used to check for nuclear and cytoplasmic fractionation, respectively. (c) Western blot of RNP complexes immunopurified from cytoplasmic fractions with anti-TIA-1 and anti-TIAR antibodies in control (Untr) and Ars-treated cells. Antibodies against TDP-43 and HuR revealed an enrichment of these two proteins in the cytoplasm in association to TIA-1 and TIAR after inducing oxidative stress. Western blots with TIAR and TIA-1 served as a positive control for immunoprecipitation. The arrowhead indicates the antibody light chain.

TDP-43-positive SGs are induced by distinct cellular insults

We studied whether other cellular insults, already reported to favour SG assembly, were also able to change TDP-43 sub-cellular distribution. We observed that TDP-43 co-localized with the SG markers TIAR and HuR in granules after 30-min exposure of NSC34 cells to heat shock (Figs 3a,b and S1) and that pre-treatment with emetine abolished the formation of such cytoplasmic structures (Fig. 3c). Recently, also the pharmacological inhibition of the ubiquitin-proteasome system was shown to induce SG assembly (Mazroui et al. 2007). We found that treatment of NSC34 cells with the specific ubiquitin-proteasome system inhibitor MG132 promoted the recruitment of TDP-43 protein into cytoplasmic foci which stained positive for the HuR marker (Fig. S1). Again, the use of the two translational inhibitors, puromycin and emetine, had opposite effects on SG formation and TDP-43 was present in granules only when cells were incubated with puromycin before treatment with MG132 (Fig. S1).

Figure 3.

 TDP-43 forms SGs in response to heat shock. Immunofluorescence images of TIAR (green) and TDP-43 (red) protein localization before (a) and after (b) exposing NSC34 cells to a heat shock for 30 min at 44°C. (c) The presence of the translational inhibitor emetine prevented SG formation. Nuclei were counter-stained by DAPI (blue). Scale bar, 10 μm.

TDP-43 protein is specifically recruited to SGs and not to P-bodies

Several RBPs have been described to participate to SG assembly in condition of environmental stress (Anderson and Kedersha 2006), although it is not clear if this recruitment is specific or common to all RBPs. To address this issue we analyzed the cell distribution of another RBP, Nova1, after inducing oxidative stress by arsenite treatment. Nova1 protein, which is an important alternative splicing factor in neurons (Jensen et al. 2000), neither changed its localization nor co-localized with the SG marker TIAR in condition of oxidative insult (Fig. 4a).

Figure 4.

 TDP-43 is a component of SGs, but not of P-bodies. (a) Distribution of TIAR (green) and Nova1 (red) proteins before (Untr) and after exposing NSC34 cells to arsenite (Ars) treatment. (b) Immunocytochemistry of NSC34 cells after 48-h co-transfection of the GFP-tagged plasmid encoding for the P-bodies marker Dcp1 (green) and the FLAG-tagged TDP-43 construct (red) before and after inducing oxidative stress. Nuclei were visualized by DAPI (blue). Scale bar, 10 μm.

We also investigated whether TDP-43 was present in P-bodies, constitutive cytoplasmic RNP complexes which control mRNA fate and degradation. NSC34 cells were transiently co-transfected with a Green Fluorescent Protein (GFP)-tagged plasmid encoding for the P-bodies marker Dcp1 (Decapping enzyme 1) and a FLAG-tagged TDP-43 construct. Recombinant TDP-43 protein did not show co-localization with Dcp1 both before and after arsenite treatment, being TDP-43 mainly nuclear in unstressed condition and being present in distinct cytoplasmic punctate granules after induction of oxidative stress (Fig. 4b).

Both RRM1 domain and C-terminal region are necessary for TDP-43 aggregation in SGs

We used distinct deletion constructs to identify the aminoacidic regions responsible for TDP-43 assembly into SGs in condition of oxidative stress (Fig. 5a).

Figure 5.

 The RRM1 domain and the C-terminal 216–315 aminoacidic region are both required for TDP-43 targeting to SGs. (a) A schematic representation of TDP-43 constructs is shown. FLAG-tagged deleted (ΔRRM1, 1–315 and ΔC) plasmids were used for transient transfection of NSC34 cells. (b–d) Immunofluorescence images of transfected NSC34 cells expressing the different recombinant TDP-43 proteins in physiological and oxidative (Ars) conditions. TIAR (green) and FLAG (red) antibodies were used for detecting SGs and transfected cells, respectively. TIAR and FLAG co-localization signals are shown (merge), while nuclei are stained in blue (DAPI). Scale bar, 10 μm.

The mutant ΔRRM1 protein, lacking the entire RRM1 domain responsible for target RNA binding, distributed in discrete intra-nuclear bodies in physiological conditions as already described (Ayala et al. 2008). After the arsenite insult was applied, the mutant TDP-43 ΔRRM1 protein failed to shuttle into the cytoplasm where TIAR-positive SGs were clearly evident (Fig. 5b).

When we used two different FLAG-tagged constructs, 1–315 and ΔC (carrying residues 1–216), with a partial and full deletion, respectively, of the TDP-43 C-terminal region, a similar sub-cellular distribution of the two truncated proteins was observed in untreated NSC34 cells. Both mutant proteins showed a variable localization, being present either exclusively in the nucleus or also in the cytoplasm (Fig. 5c and d), but their response to arsenite treatment was completely different. In fact, the TDP-43 1–315 protein formed cytoplasmic foci, which stained positive for the SG marker TIAR, in about 35% of the transfected cells (Fig. 5c), while the ΔC truncated protein was not able to assemble into SGs following the cellular insult (Fig. 5d). These data indicate that both TDP-43 RRM1 and C-terminal region, in particular the aminoacidic interval spanning from residue 216 to 315 (Fig. 5a), are necessary for the recruitment of TDP-43 into SGs in conditions of oxidative stress.

TDP-43 is neither an essential component of SGs nor a neuroprotective factor in stress condition

To further assess the role of TDP-43 in SG assembly and in stress response, we transfected the FLAG-tagged full-length TDP-43 construct in NSC34 cells and analyzed its capacity to induce SGs in the absence of any environmental insult. The over-expression of the RBPs TIA-1 and TIAR was previously described to promote SG assembly in physiological conditions (Gilks et al. 2004) and this was also confirmed in our motoneuronal cell model (Fig. S2). On the contrary, we found that the over-expression of TDP-43 protein was not sufficient per se to induce SG formation. As we have already observed for the endogenous TDP-43 protein, FLAG- and TIAR-positive SGs only formed following treatment with arsenite (Fig. S2).

To address the issue of TDP-43 function in oxidative stress response, we silenced its expression in NSC34 cells by using a specific siRNA duplex. The efficiency of TDP-43 knock-down was evaluated by western blot analysis at two different time points, obtaining a 70% and a 90% reduction at 72 and 96 h post-transfection, respectively (Fig. S3). When cells were treated with arsenite, we observed that TIAR-positive SGs were able to form in TDP-43-depleted cells (Fig. 6a), as well as in cells transfected with an irrelevant control siRNA (Fig. 6b). We also evaluated whether the lack of TDP-43 would affect cell survival in response to oxidative stress in a time-course assay. The viability of TDP-43-knocked down cells, exposed to arsenite for 30, 60, 90 and 120 min, was similar to the control siRNA-transfected cells (Fig. S4).

Figure 6.

 TDP-43 is not necessary for SG formation. Double immunofluorescence staining of NSC34 cells knocked-down for TDP-43 (siTDP-43) and, as a control, for firefly luciferase (siCtrl) after treatment with 0.5 mM arsenite for 30 min. TIAR (green) and TDP-43 (red) antibodies were used. Arrows indicate TDP-43-depleted cells. Merged images are shown and nuclei visualized by DAPI staining (blue). Scale bar, 10 μm.

TDP-43-positive inclusions in human ALS motor neurons do not contain SG markers

As ALS-affected motor neurons show abnormal cytoplasmic protein aggregates, degenerating mitochondria and an increased level of reactive oxygen species, we investigated whether SGs were present in spinal cord tissue of ALS patients and whether the described pathological TDP-43 inclusions also contained SG markers. We analyzed the sub-cellular distribution of the RBPs TIAR and HuR in autoptic spinal cord from three patients affected by sporadic ALS. By fluorescent immunostaining assays we observed the presence of TDP-43-positive inclusions in many affected motor neurons in all the analyzed patients, with either filamentous (Fig. 7) or compact round shapes (data not shown). Such large cytoplasmic inclusions did not stain positive for the two SG markers TIAR and HuR (Fig. 7a and b). Additionally, we observed a great intra-patient variability in the sub-cellular distribution of the TDP-43 protein. In some motor neurons we observed a physiological localization of TDP-43 showing a main nuclear staining (Fig. 7c), while in others TDP-43 appeared completely mislocalized in the cytoplasm although no large filamentous or round inclusions were evident (Fig. 7d). Interestingly in these cases TDP-43 distributed in small and discrete cytoplasmic granules which did not overlap with either HuR or TIAR RBPs (see enlargements in Fig. 7c and d).

Figure 7.

 TDP-43 inclusions do not stain positive for SG markers in ALS brain tissues. Human spinal cord tissue of patients with ALS were stained with the SG markers TIAR (a, red) or HuR (b, red) together with TDP-43 (green). Merged images (merge) show no distribution of TIAR and HuR RBPs in TDP-43-positive filamentous inclusions. Enlargements of the indicated areas are shown. Further magnification on the z-plan of the granules pointed by the arrow is shown in the inset. (c) Some motor neurons showed a normal distribution of TDP-43 (green) and HuR (red) proteins in the nucleus and a minor localization in discrete granules in the cytoplasm with no evident co-localization signals (merge and inset images). (d) Immunofluorescence image of an affected motor neuron where TDP-43 (green) was completely mislocalized in the cytoplasm in a granule-like distribution, without forming inclusions. TIAR staining (red) does not overlap TDP-43 signals (merge and inset images). The asterisk (*) indicates lipofuscin granules. Scale bar, 20 μm.

Discussion

In this paper, we proved that TDP-43 is capable to respond to an environmental insult by assembling into stress granules (SGs), cytoplasmic ribonucleoprotein foci which sequester mRNAs, several RBPs and stalled translation initiation complexes to temporarily arrest protein synthesis as a protective response to cellular stress.

The genetic findings in ALS patients of causative mutations in three proteins involved in RNA processing, namely Senataxin, TDP-43 and Fusion/Translocated in LipoSarcoma (FUS/TLS) (Chen et al. 2004; Sreedharan et al. 2008; Kwiatkowski et al. 2009; Vance et al. 2009), have focused the attention on the complex molecular mechanisms regulating gene expression at post-transcriptional level as potential pathogenic clues. Post-transcriptional control of mRNA fate is known to play an important role both during the development of the nervous system and for the maintenance of neural activities in the adult brain. RBP-mediated regulatory mechanisms allow a precise spatio-temporal control of mRNA translation, associated to transport and subcellular compartmentalization of mRNAs in dendrites and axons (Besse and Ephrussi 2008), so that disruption of such activities is supposed to severely impair neuronal cell metabolism.

So far, TDP-43 functions have been mainly associated to alternative splicing processes and transcriptional activities (Bose et al. 2008; Buratti and Baralle 2008). However, TDP-43 protein has been recently demonstrated to have multiple roles in the regulation of mRNA fate in neuronal cells, such as transcript stabilization and activity-dependent transport to dendrites (Strong et al. 2007; Wang et al. 2008). In this view, the sequestration of TDP-43 in pathological aggregates is supposed to determine a loss of function of the protein with severe consequences on mRNA metabolism and post-transcriptional regulation of gene expression.

Our results show that in the motoneuron-like NSC34 cells different types of environmental insults, ranging from oxidative stress to proteasome inhibition and heat shock, are able to induce the assembly of TIAR- and HuR-positive SGs, a subset of which also contains TDP-43 protein. These data further indicate that TDP-43 has also a role in the control of mRNA fate in the cytoplasmic compartment. It is interesting that also FUS/TLS, another RBP causative of 5% of familial ALS cases (Lagier-Tourenne and Cleveland 2009), forms SGs in conditions of oxidative stress (Andersson et al. 2008). Indeed, several RBPs, enzymes and cytoskeletal elements have already been described to be components of SGs (Anderson and Kedersha 2008; Tsai et al. 2009). However, we demonstrated that not every RBP is necessarily included in these structures after environmental stress. In fact Nova1, a neuron-specific splicing factor essential for the development of the motor system and for the survival of motor neurons (Jensen et al. 2000), did not change its sub-cellular distribution in response to arsenite insult. This observation is particularly interesting as both Nova1 and TDP-43 are mainly nuclear proteins regulating alternative splicing of pre-mRNA, but also shuttle actively to the cytoplasm (Ayala et al. 2008; Ratti et al. 2008; Wang et al. 2008). Therefore, TDP-43 RBP is specifically recruited to SGs but is not an essential component. This was shown both by over-expression and gene silencing experiments and by the fact that TDP-43-positive SGs are present only in a subset of SG-forming NSC34 cells. Moreover, we demonstrated that TDP-43 is not able to influence cell viability following stress conditions, supporting the idea that TDP-43 participates to regulatory mechanisms of translational arrest, but it is not one of the promoting factors.

Our data also show that TDP-43 is not involved in targeting bound transcripts to the degradation machinery of P-bodies, but rather in provisionally silencing them in presence of a toxic insult. TDP-43 is known to specifically recognize and bind repeated (UG)n motifs and this may help explain why a low abundance of TDP-43-positive SGs is observed, as it likeky reflects the low frequency of (UG)n-containing mature mRNA. Few target mRNA have been identified and validated so far for TDP-43 (Buratti and Baralle 2008; Wang et al. 2008) and the future identification of all its targets is expected to unravel the biological role of this RBP in neuronal cells and, at the same time, its pathomechanism in neurodegenerative diseases.

In our attempt to define the aminoacidic region responsible for TDP-43 assembly into SGs, we found that the TDP-43 protein lacking the RRM1 domain failed to shuttle from the nucleus to the cytoplasm after stress. The RRM1 domain is responsible for target mRNA binding, in particular for the recognition of UG repeat motifs (Buratti and Baralle 2001) so that its disruption may determine a defective mRNA binding and protein function. In fact, in physiological condition TDP-43 ΔRRM1 mutant distributes in the nucleus in abnormal granular structures that may be associated with changes in chromatin distribution of the ΔRRM1 protein (Fig. 5c; Ayala et al. 2008).

An active and efficient shuttling from the nucleus to the cytoplasm and vice-versa is necessary for the proper activity of many RBPs, including FUS/TLS and the Embryonic Lethal Abnormal Vision (ELAV) family member HuR. Like TDP-43, FUS/TLS and HuR show a main nuclear distribution in physiological conditions, but they also function as mRNA transport and stabilizing factors in the cytoplasm (Fujii and Takumi 2005; Hinman and Lou 2008). Interestingly, in familial ALS, mutant FUS/TLS protein shows an abnormal redistribution in the cytoplasm (Kwiatkowski et al. 2009; Vance et al. 2009) similarly to what has been observed with HuR in ALS animal and cellular models (Lu et al. 2007), suggesting a potential perturbation of their original function.

A mislocalization of TDP-43 protein from the nucleus to the cytoplasm already occurs in physiological conditions with mutant forms lacking both the C-terminus and the N-terminal NLS (Nuclear Localization Signal) region (Ayala et al. 2008; Johnson et al. 2008; Winton et al. 2008; Nonaka et al. 2009). More importantly, we have found that, besides the RRM1 domain, also the selective lack of 100 aminoacids (216–315) in the C-terminal region determines the failure of TDP-43 to assemble into SGs. This glycine-rich C-terminal region is highly conserved along phylogenesis and is supposed to be involved in protein-protein interactions. However, the ability of TDP-43 to aggregate into SGs seems not to be related to its interaction with heterogeneous RNP A/B proteins, essential for its splicing activity, as the binding region maps to residues 321–366 (D’Ambrogio et al. 2009). Interestingly, the C-terminal domain represents the mainly mutated region in ALS patients, although distinct mutant TDP-43 proteins (A382T, Q331K and M337V) distribute normally in the nucleus (Sreedharan et al. 2008), suggesting that alteration in TDP-43 sub-cellular localization is determined by multiple aminoacidic changes and/or additional contributing factors.

Several pathogenic mechanisms have been demonstrated to trigger motor neuron death in ALS, although the primary or secondary role of these events in the pathogenesis of the disease is not clear yet, and the early causative processes still need to be clarified. As these mechanisms include oxidative stress, mithocondrial defects, protein aggregation, and proteasome impairment, we investigated if SGs may be implicated in ALS pathogenesis. In human ALS motor neurons we failed to detect TIAR- and HuR-positive granules or inclusions co-localizing with TDP-43 and, also in conditions of a main physiological TDP-43 distribution in the nucleus, we did not observe overlapping signals with the SG markers TIAR and HuR in the cytoplasm. A previous paper reported the presence of TIA-1 and PolyA-Binding Protein 1 (PABP-1), another RBP included in SGs, in the RNA-positive basophilic inclusions from patients presenting with adult-onset atypical motor neuron disease (Fujita et al. 2008), confirming our hypothesis that translational arrest mechanisms are somehow involved in neuronal death in vivo. The observation that SG structures are not associated only to environmental insults in vitro, but may form and be detectable also in disease conditions, has already been reported in animal models of brain ischemia and sciatic nerve axotomy (DeGracia et al. 2008; Moisse et al. 2009) as well as in tumors exposed to radiation-induced hypoxia (Moeller et al. 2004).

One possible explanation to the fact that in ALS human motor neurons SGs are not present is that such macromolecular complexes may assemble as a very early response to an environmental stress and that at the endstage of the neurodegenerative process other compensative mechanisms may have occurred. On the other hand, we may also hypothesize that in ALS affected tissues stressful processes are progressive and not acute enough to provoke SG formation, like in ischemia and axotomy conditions. We have experimental evidence that TDP-43 does not form SGs, but distributes diffusely in the cytoplasm when treating NSC34 cells with milder doses of arsenite for prolonged periods of time to mimic a chronic insult (unpublished results). Our findings, together with the observed cytoplasmic mislocalization of HuR in mutant Superoxide dismutase 1 (SOD1) transgenic mice at very early stages of the disease (Lu et al. 2007), support the idea that impairment of post-transcriptional regulatory mechanisms, including mRNA stabilization and translation, may be actively involved in ALS pathogenesis and/or progression. However, the reason of such a specific TDP-43, and not HuR, pathological aggregation in human ALS brain tissues needs further investigation. Although preliminary data show the absence of FUS/TLS-positive cytoplasmic inclusions in sporadic ALS patients (Kwiatkowski et al. 2009; Vance et al. 2009), a full comprehension of the potential interplay of this protein with TDP-43 and HuR will help elucidating the role of RBP-mediated regulation of gene expression in neurodegeneration and motor neuron diseases.

Acknowledgements

We thank prof. F.E. Baralle for critically reading the manuscript, E. Giovannini for her technical help and G. Bassell. This work was financially supported by the Italian Ministry of Health (Malattie Neurodegenerative, ex Art.56, n.533F/N1), Fondazione Cariplo (Grant n.2008.2307) and a donation of Peviani Family. EB is supported by Telethon and Eurasnet.

Ancillary