USP10 Inhibits Aberrant Cytoplasmic Aggregation of TDP-43 by Promoting Stress Granule Clearance

ABSTRACT

TAR DNA-binding protein 43 (TDP-43) is a causative factor of amyotrophic lateral sclerosis (ALS). Cytoplasmic TDP-43 aggregates in neurons are a hallmark pathology of ALS. Under various stress conditions, TDP-43 localizes sequentially to two cytoplasmic protein aggregates, namely, stress granules (SGs) first and then aggresomes. Accumulating evidence suggests that delayed clearance of TDP-43-positive SGs is associated with pathological TDP-43 aggregates in ALS. We found that ubiquitin-specific protease 10 (USP10) promotes the clearance of TDP-43-positive SGs in cells treated with proteasome inhibitor, thereby promoting the formation of TDP-43-positive aggresomes, and the depletion of USP10 increases the amount of insoluble TDP-35, a cleaved product of TDP-43, in the cytoplasm. TDP-35 interacted with USP10 in an RNA-binding-dependent manner; however, impaired RNA binding of TDP-35 reduced the localization in SGs and aggresomes and induced USP10-negative TDP-35 aggregates. Immunohistochemistry showed that most of the cytoplasmic TDP-43/TDP-35 aggregates in the neurons of ALS patients were USP10 negative. Our findings suggest that USP10 inhibits aberrant aggregation of TDP-43/TDP-35 in the cytoplasm of neuronal cells by promoting the clearance of TDP-43/TDP-35-positive SGs and facilitating the formation of TDP-43/TDP-35-positive aggresomes.

INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive lethal motor neuron disease, characterized by a loss of motor neurons in the spinal cord, motor cortex, and brainstem (1). Frontotemporal lobar degeneration (FTLD), a frequent cause of dementia, is characterized by the loss of neurons in the frontal and/or temporal lobes of the brain (2). TAR DNA-binding protein 43 (TDP-43) is a major causative protein of familial and sporadic ALS and FTLD with TDP-43 pathology (FTLD-TDP) (3, 4). TDP-43 is a nuclear RNA-binding protein; however, TDP-43 generates cytoplasmic ubiquitinated TDP-43 inclusions in neurons of ALS and FTLD-TDP patients, and these TDP-43 inclusions are tightly linked to the development of ALS and FTLD-TDP. How these pathogenic cytoplasmic TDP-43 inclusions in ALS and FTLD-TDP neurons are generated remains to be elucidated.

Extrinsic and intrinsic stresses, such as oxidative stress, heat shock, viral infection, and proteasome dysfunction, increase the amount of ubiquitinated proteins and then induce ubiquitinated protein aggregates with neurotoxicity (5). Thus, the aberrant accumulation of ubiquitinated protein aggregates in cells and tissues is a causative factor of many neurodegenerative diseases, including ALS and FTLD-TDP (6, 7). In response to these various stresses, cells form two types of protein aggregates in the cytoplasm, namely, stress granules (SGs) and aggresomes.

SGs are stress-inducible cytoplasmic ribonucleoprotein aggregates containing many mRNAs and RNA-binding proteins, including TDP-43 (8, 9). SG-inducing stresses include heat shock, oxidants, virus infection, and proteasome inhibitors (PIs) (10). These stresses induce the disassembly of polysomes, and then disassembled mRNAs and RNA-binding proteins promote SG formation (11). Thus, SGs contain many ribosomal proteins and translation-associated proteins, such as eIF-4E and poly(A)-binding protein (PABP). After stress disappears, intact proteins and mRNAs in SGs reassemble polysomes to restart translation for recovery (12). SGs have stress-protective functions. For instance, SG formation inhibits stress-induced reactive oxygen species (ROS) production and ROS-dependent apoptosis (13, 14).

Aggresomes are stress-inducible cytoplasmic protein aggregates that contain many ubiquitinated proteins (15). They are formed in the perinuclear cytoplasmic region called the microtubule-organizing center (MTOC). Several stress agents, such as oxidants, heat shock, and PIs, induce aggresomes in cultured cells. These stresses increase the amount of ubiquitinated proteins. Then, these ubiquitinated proteins are captured by ubiquitin (Ub) receptor protein p62, and such p62-bound ubiquitinated proteins are transported to the perinuclear MTOC to form aggresomes (16). Aggresomes are tightly linked to autophagy, and several aggresome components are degraded by p62-dependent selective autophagy (aggrephagy) (17). Histone deacetylase 6 (HDAC6) plays a critical role in aggresome formation by promoting the transport of p62 aggregates to the MTOC in a microtubule-dependent manner (18, 19). In addition, HDAC6 promotes the formation of SGs in a microtubule-dependent manner (20). Thus, HDAC6 is involved in the cross talk between SGs and aggresomes.

Two groups of proteins, i.e., Ras-GAP SH3 domain-binding protein 1 (G3BP1)/G3BP2 and Ub-specific protease 10 (USP10), regulate SG and aggresome formation, respectively (19, 21, 22). G3BP1 and G3BP2 are RNA-binding proteins and are well known to initiate SGs by forming homooligomers and heterooligomers. The depletion of the expression of G3BP1 and G3BP2 in cultured cells prominently reduces SG formation (21, 22). USP10 is a binding partner of G3BP1/G3BP2 (21, 22). Interestingly, USP10 acts as an initiator of aggresome formation (19). While USP10 depletion in cultured cells reduces aggresome formation, the overexpression of USP10 together with a large amount of ubiquitination-prone proteins induces aggresome formation without stress. To induce aggresomes, USP10 interacts with p62, and the interaction augments ubiquitinated protein aggregation and aggresome formation (19, 23). These results suggest that G3BP1/G3BP2 and USP10 cooperatively control a stress-protective response by forming SGs and aggresomes.

SGs contain several neuropathogenic RNA-binding proteins, including TDP-43 and FUS (8). Intriguingly, an increasing number of studies have suggested that pathological aggregation of several RNA-binding proteins, including TDP-43, in ALS is initiated in abnormal types of SGs (aberrant SGs) (9, 24). Aberrant SGs have different characteristics, compared with normal SGs. For example, aberrant SGs have delayed SG clearance, compared with normal SGs (25). In addition, aberrant SGs contain more ubiquitinated proteins than normal SGs (26). Although some proteins, including valosin-containing protein (VCP) and p62, have been shown to promote SG clearance (27, 28), the mechanisms underlying SG clearance and the formation of aberrant SGs have not been fully elucidated.

In the present study, we investigated how abnormal cytoplasmic aggregation of TDP-43 is induced in cells treated with PI. Proteasome inhibition of cultured cells generated a 35-kDa C-terminal fragment of TDP-43 (TDP-35), which has RNA recognition motifs (RRMs) without a nuclear localization signal (NLS) (29), and TDP-35/TDP-43 sequentially localized in SGs and aggresomes. We found that USP10 promotes the clearance of TDP-43/TDP-35-positive SGs, thereby promoting the formation of TDP-43/TDP-35-positive aggresomes, and the depletion of USP10 increased the amount of insoluble cytoplasmic TDP-35 but not TDP-43. TDP-35 interacted with G3BP1 and USP10, and the interaction was well correlated with the localization of TDP-35 in SGs and aggresomes. Decreasing the RNA-binding activity of TDP-35 abolished its interaction with USP10 and G3BP1, resulting in the generation of aberrant USP10/G3BP1-negative TDP-35 aggregates in the cytoplasm. Furthermore, TDP-43/TDP-35 aggregates in ALS motor neurons were mostly USP10 negative.

The present findings suggest that USP10 inhibits the formation of aberrant TDP-43/TDP-35 aggregates in SGs in neuronal cells by promoting SG clearance and aggresome formation, but reduced RNA-binding activity of TDP-43/TDP-35 allows an escape from this aggregate protection mechanism and may induce pathological aggregates in ALS neurons.

RESULTS

Proteasome inhibition sequentially induces TDP-43-positive SGs and aggresomes.

To examine the roles of SGs and aggresomes in TDP-43 metabolism under stress conditions, we first characterized the localization of TDP-43 in SGs and aggresomes in 293T cells treated with PI MG-132 or bortezomib (BTZ). At 4 h after MG-132 treatment, 293T cells formed SGs that were detected with three SG marker proteins (G3BP1, USP10, and PABP), and the number of SG-positive cells then decreased to 12 h (Fig. 1A and B; also see Fig. S1A in the supplemental material) (30). On the other hand, aggresomes were initially detected at 8 h after PI treatment, as one large aggregate stained with the aggresome marker protein p62 in the perinuclear cytoplasmic regions (19), and the number of aggresome-positive cells increased until 12 h (Fig. 1A and B). Similar sequential formations of SGs and aggresomes were observed in 293T cells treated with BTZ (Fig. 1B). In addition, SGs and aggresomes with similar kinetics were induced in a neuroblastoma cell line, Neuro2a (Fig. 1C and D). These results suggested that SGs and aggresomes are coordinately induced in neuronal and nonneuronal cells treated with PIs. Interestingly, in MG-132-treated cells, two SG-localized proteins, G3BP1 and USP10, were enriched around the aggresome rather than in the center of the aggresome (Fig. 1A and C).

FIG 1.

TDP-43 is localized in SGs and then aggresomes. (A) 293T cells were treated with 5 μM MG-132 for the indicated times, and then the cells were fixed and stained with anti-USP10 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). The scale bar indicates 10 μm (A, C, E, and G). (B) The proportions of cells containing SGs (USP10 positive) and aggresomes (p62 positive) under the treatment of 5 μM MG-132 or 1 μM BTZ (BTZ) are presented as the mean ± SD (n = 4). Cells with one large p62-positive aggregate (more than 15 μm² in size) in the perinuclear region with nuclear deformity were considered aggresome-positive cells. (C) Neuro2a cells were treated with 5 μM MG-132 for the indicated times, and the cells were stained with anti-G3BP1 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). (D) The proportions of cells containing SGs (G3BP1 positive) and aggresomes (p62 positive) are presented as the mean ± SD (n = 3). (E and G) 293T cells (E) or Neuro2a cells (G) were treated with 5 μM MG-132 for the indicated times, and the cells were stained with anti-TDP-43 antibodies (green) with either anti-G3BP1 antibodies or anti-p62 antibodies (red), together with Hoechst 33258 (blue). (F and H) The proportions of cells containing TDP-43/G3BP1-positive SGs or TDP-43/p62-positive aggresomes in 293T cells (F) or Neuro2a cells (H) are presented as the mean ± SD (n = 3).

TDP-43 is a causative protein in familial and sporadic ALS and FTLD-TDP (3, 4). TDP-43 is an RNA-binding protein, and aberrant cytoplasmic aggregates of TDP-43 in neurons play a pathognomonic role in the development of ALS and FTLD-TDP. Intriguingly, TDP-43 has been detected in SGs and aggresomes in cultured cells under several stress conditions (31–33). Thus, we next examined the localization of TDP-43 in SGs and/or aggresomes in 293T and Neuro2a cells after PI treatment. Prior to MG-132 treatment, TDP-43 was mainly localized in the nucleus and weakly localized in the cytoplasm of 293T and Neuro2a cells. After MG-132 treatment, a part of TDP-43 was detected in SGs at 4 h and in aggresomes at 8 to 12 h (Fig. 1E to H). We also detected TDP-43 in MG-132-induced SGs and aggresome-like aggregates of primary neuron-enriched cells prepared from the embryonic rat brain (see Fig. S2). These results suggested that TDP-43 is localized in SGs and then aggresomes in three types of cells, including primary neurons, after PI treatment.

G3BP1 is a critical factor of TDP-43-positive SG formation induced by PI treatment.

G3BP1 is an initiator of SG formation in cultured cells treated with various stress stimuli (22, 34–36). Depletion of both G3BP1 and G3BP2 expression in U2-OS cells inhibits the formation of TDP-43-positive SGs by sodium arsenite treatment (31). However, the function of G3BP1 in the formation of TDP-43-positive SGs with PI treatment in 293T cells has not been clarified. To check this, we transfected G3BP1 small interfering RNA (siRNA) (G3BP1-knockdown [KD]) into 293T cells, and cells were treated with MG-132 for 4 h. G3BP1-KD in MG-132-treated cells prominently decreased the number of cells with TDP-43-positive SGs (Fig. 2A to C). These results showed that G3BP1 promotes the formation of SGs after PI treatment, and the induced SGs recruit TDP-43.

FIG 2.

G3BP1-KD reduces TDP-43-positive SG formation induced by PI. (A) 293T cells were transfected with nontargeting siRNA (NT) or G3BP1-1-siRNA (FlexiTube siRNA of Qiagen) for 48 h, and whole-cell extracts were characterized by Western blotting using anti-G3BP1, anti-USP10, and anti-β-actin antibodies. (B) The indicated cells were treated with 5 μM MG-132 or dimethyl sulfoxide (DMSO) for 4 h. Then, cells were stained with anti-TDP-43 antibodies (green), anti-PABP antibodies (red), and Hoechst 33258 (blue). The scale bar indicates 10 μm (B, E, and H). (C) The proportions of cells containing TDP-43/PABP-positive SGs are presented as the mean ± SD (n = 3). Statistical significance was determined by a two-tailed unpaired t test (C and I). (D) A schematic illustration of the structure of the TDP-43 and TDP-35 proteins used in this study. (E) Neuro2a or 293T cells were transfected with the GFP-TDP-35 plasmid, and then cells were stained with anti-G3BP1, anti-PABP, or anti-USP10 antibodies (red) and Hoechst 33258 (blue). (F) 293T cells were transfected with nontargeting siRNA (NT), G3BP1-1-siRNA, G3BP2-siRNA, or G3BP1-1 plus G3BP2 siRNAs (G1/G2) for 24 h and further transfected with the GFP-TDP-35 plasmid for 24 h. Whole-cell extracts were characterized by Western blotting using anti-G3BP1, anti-G3BP2, and anti-β-actin antibodies. (G) The proportions of cells containing GFP-TDP-35/PABP-positive SGs are presented as the mean ± SD (n = 4). Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test. (H) 293T cells were transfected with the FLAG-G3BP1 plasmid together with the GFP-TDP-35 plasmid. The cells were then fixed and stained with anti-G3BP1 antibodies (red) and Hoechst 33258 (blue). (I) The proportions of cells containing GFP-TDP-35-positive SGs (n = 3) and the sizes of GFP-TDP-35-positive SGs (n = 30) are presented as the mean ± SD. The levels of statistical significance are as follows: **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

To further study the G3BP1 function in TDP-43-positive SG formation, we characterized a fusion protein of TDP-43 with a green fluorescent protein (GFP) (GFP-TDP-43); however, GFP-TDP-43 was not detected in G3BP1-positive SGs with MG-132 treatment, due to the exclusive nuclear localization of GFP-TDP-43 (see Fig. S3A). Interestingly, the overexpression of GFP-TDP-43 induced G3BP1-positive SGs; however, again, these SGs were not colocalized with GFP-TDP-43. Thus, we next used GFP-TDP-35 (amino acids 86 to 414 of TDP-43), which has RRM1 and RRM2 but not the NLS (Fig. 2D). Without MG-132 treatment, GFP-TDP-35 was detected in both the nucleus and the cytoplasm and induced GFP-TDP-35-positive SGs with three SG marker proteins (G3BP1, PABP, and USP10) in approximately 20% of Neuro2a and 293T cells (Fig. 2E). In MG-132-treated cells, 28.3% of untagged TDP-35 localized to SGs, compared with 6.0% of untagged TDP-43, confirming that TDP-35 localized to SGs more than TDP-43 (see Fig. S3B). The number of cells with GFP-TDP-35-positive SGs was decreased by G3BP1-KD or G3BP2-KD (Fig. 2F and G; also see Fig. S3C to E). G3BP2 is a homologue of G3BP1 (21). G3BP2-KD has been shown to decrease the number of cells with SGs in cultured cells, but the reduction is less than that of G3BP1 (21). Furthermore, overexpression of G3BP1 increased the number of cells with GFP-TDP-35-positive SGs (Fig. 2H and I). This result is consistent with a previous report that G3BP overexpression induces SGs in a dose-dependent manner (37). Taken together, these results indicate that G3BP1 and G3BP2 induce the formation of SGs in 293T cells and the induced SGs recruit TDP-35 much more efficiently than TDP-43.

Treatment of cells with MG-132 induced the formation of SGs followed by the formation of aggresomes (Fig. 1B and D). Next, we examined the involvement of G3BP1 in the formation of aggresomes. Interestingly, G3BP1-KD not only decreased the number of cells with SGs at 4 h after MG-132 treatment (Fig. 3A and B) but also decreased the number of cells with aggresomes at 12 h after MG-132 treatment (Fig. 3C and D). These results indicate that G3BP1 promotes the formation of both TDP-43-positive SGs and TDP-43-positive aggresomes in 293T cells treated with MG-132.

FIG 3.

G3BP1-KD reduces the formation of TDP-43-positive SGs and aggresomes. (A) 293T cells were transfected with nontargeting siRNA (NT) or G3BP1-siRNA (G3BP1-1 or G3BP1-2) for 48 h, and cells were treated with 5 μM MG-132 for 4 h. The cells were then fixed and stained with anti-TDP-43 antibodies (green), anti-PABP antibodies (red), and Hoechst 33258 (blue). The scale bar indicates 10 μm (A and C). (B) The proportions of cells containing TDP-43/PABP-positive SGs are presented as the mean ± SD (n = 3). Statistical significance was determined by a two-tailed unpaired t test (B and D). (C) 293T cells were transfected with nontargeting siRNA (NT) or G3BP1-siRNA (G3BP1-1 or G3BP1-2) for 48 h, and cells were treated with 5 μM MG-132 for 12 h. The cells were then fixed and stained with anti-TDP-43 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). (D) The proportions of cells containing TDP-43/p62-positive aggresomes are presented as the mean ± SD (n = 3). The levels of statistical significance are as follows: *, P < 0.05; **, P < 0.01.

USP10 promotes the formation of TDP-43-positive aggresomes by facilitating the clearance of SGs in PI-treated cells.

USP10 is a binding protein of G3BP1 and G3BP2 (21, 22). USP10 plays a critical role in the formation of MG-132-induced aggresomes in cultured cells (19). Thus, we next examined whether USP10 plays a role in the formation of TDP-43-positive aggresomes using USP10-KD (Fig. 4A and B). MG-132 treatment for 12 h induced one large cytoplasmic TDP-43/p62-positive aggresome in USP10-competent (USP10-wild-type [WT]) cells, but the number of cells with TDP-43-positive aggresomes was decreased by USP10-KD (Fig. 4A and B). In MG-132-treated USP10-KD cells, TDP-43 was diffusely detected throughout the cytoplasm. Taken together, these results suggest that USP10 plays a critical role in the formation of TDP-43-positive aggresomes in MG-132-treated cells and the depletion of USP10 induces aberrant cytoplasmic TDP-43 localization.

FIG 4.

USP10-KD reduces TDP-43-positive aggresome formation induced by PI. (A) 293T cells were infected with lentivirus encoding control nontargeting shRNA (NT) or USP10-shRNA (USP10-1 or USP10-3), and the cells were cultured in the presence of puromycin. These cells were treated with 5 μM MG-132 or DMSO for 12 h. The cells were then fixed and stained with anti-TDP-43 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). The scale bar indicates 10 μm. (B) The proportions of cells containing TDP-43/p62-positive aggresomes are presented as the mean ± SD (n = 3). Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test (B and C). (C to E) Control (NT) and USP10-KD (USP10-1 or USP10-3) cells were treated with 5 μM MG-132 or DMSO for 12 h, and then whole-cell extracts (C), cytoplasmic/nuclear extracts (D), or NP-40-soluble/insoluble extracts (E) were characterized by Western blotting using the indicated antibodies. The amounts of TDP-43 and TDP-35 were quantified by densitometric scanning of the corresponding bands of the Western blot (C). The levels of statistical significance are as follows: NS, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Western blotting detected full-length TDP-43 protein and TDP-43 derivative (referred to as TDP-35 in this study) in 293T cells treated with MG-132, and the amount of TDP-35, but not TDP-43, was increased by MG-132 treatment and further increased by USP10-KD (Fig. 4C). TDP-35 is generated by protease-mediated cleavage of the N-terminal fragment of TDP-43 containing NLS (Fig. 2D) (29), and TDP-35 corresponds to GFP-TDP-35, which we used to characterize the formation of SGs (Fig. 2E to I). In the steady state, TDP-43 and TDP-35 were localized mainly in the nucleus (Fig. 4D). The treatment of USP10-KD and WT cells with MG-132 decreased the amount of TDP-43 and TDP-35 in the nucleus and increased the amount of TDP-35, but not TDP-43, in the cytoplasm, and the increase of TDP-35 in USP10-KD cells was greater than that in USP10-WT cells (Fig. 4D). In addition, the treatment of USP10-KD cells with MG-132 increased the amount of insoluble TDP-35 but not insoluble TDP-43; however, the increase was not observed in USP10-WT cells (Fig. 4E). These results indicate that USP10 decreases the amount of insoluble TDP-35 in the cytoplasm of cells treated with MG-132 for 12 h, which is associated with the formation of TDP-43-positive aggresomes.

Several lines of evidence indicate that the accumulation of ubiquitinated proteins in SGs delays SG clearance and leads to persistent aggregation of SG-associated proteins, such as TDP-43 (24, 26). Building on the findings of a previous study that treatment with a 70-kDa heat shock proteins (HSP70) inhibitor induced the formation of aberrant SGs containing a large amount of Ub (26), we examined the activity of USP10-KD relative to the amount of Ub and TDP-43 in SGs of cells treated with both MG-132 and an HSP70 inhibitor (VER-155008 [VER]) for 8 h (Fig. 5). Treatment with MG-132 for 8 h decreased the number of SG-positive cells, compared to 4 h treatment (Fig. 1A to D). Thus, with treatment with MG-132 for 8 h, SG clearance has already started. Treatment of USP10-WT cells with both VER and MG-132 for 8 h induced G3BP1-positive, Ub-positive, and TDP-43-positive SGs, and their numbers were significantly increased by USP10-KD (Fig. 5B to F). Western blotting showed that treatment of USP10-WT cells with both VER and MG-132 increased the amount of Ub but not TDP-43/TDP-35, and the amount was further increased by USP10-KD (Fig. 5A). These results indicate that USP10-KD induces aberrant SGs containing large amounts of ubiquitinated proteins and TDP-43/TDP-35 in cells treated with MG-132 and VER.

FIG 5.

USP10-KD promotes the formation of Ub-positive SGs induced by MG-132/HSP70 inhibitor. (A) 293T cells were transfected with nontargeting siRNA (NT) or USP10-siRNA (USP10-1 or USP10-2) for 48 h and treated with 40 μM VER or DMSO in the presence of 5 μM MG-132 for 8 h. Whole-cell extracts were characterized by Western blotting using anti-USP10, anti-G3BP1, anti-Ub, anti-TDP-43, and anti-β-actin antibodies. (B and E) Control (NT) and USP10-KD (USP10-1 or USP10-2) cells were treated with 40 μM VER or DMSO in the presence of 5 μM MG-132 for 8 h. The cells were fixed and stained with anti-G3BP1 (red) antibodies together with either anti-Ub antibodies (B) or anti-TDP-43 antibodies (E) (green). Nuclei were counterstained using Hoechst 33258 (blue). Colocalization of Ub (B) or TDP-43 (E) (green) with G3BP1 (red) was evaluated by a line profile analysis. The scale bar indicates 10 μm (B and E). (C, D, and F) The proportions of cells containing SGs with G3BP1 (C), Ub/G3BP1 (D), or TDP-43/G3BP1 (F) are presented as the mean ± SD (n = 3). Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test (C, D, and F). The levels of statistical significance are as follows: *, P < 0.05; ***, P < 0.001.

Next, we examined whether G3BP1 and USP10 regulate the formation of SGs and aggresomes in human primary neurons derived from induced pluripotent stem cells (iPSCs). More than 95% of the iPSC-derived cells were microtubule-associated protein 2 (MAP-2)-positive neurons (see Fig. S4). The treatment of iPSC-derived neurons with MG-132 sequentially increased the number of cells with SGs and aggresomes (Fig. 6A and B), and these SGs and aggresomes were colocalized with TDP-43 (Fig. 6C and D). The KD of G3BP1 and USP10 in iPSC-derived neurons decreased the numbers of cells with TDP-43-positive SGs and TDP-43-positive aggresomes, respectively (Fig. 6E to J). These results indicate that G3BP1 and USP10 control the formation of TDP-43-positive SGs and TDP-43-positive aggresomes in primary neurons.

FIG 6.

G3BP1 and USP10 regulate the formation of TDP-43-positive SGs and aggresomes in iPSC-derived neurons. (A) iPSC-derived neurons were treated with 2 μM MG-132 for the indicated times, and then the cells were fixed and stained with anti-G3BP1 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). The scale bar indicates 10 μm (A, C, G, and J). (B) The proportions of cells containing SGs (G3BP1-positive) and aggresomes (p62-positive) under treatment with 2 μM MG-132 are presented as the mean ± SD (n = 3). (C and D) iPSC-derived neurons were treated with 2 μM MG-132 for the indicated times, and the cells were stained with anti-TDP-43 antibodies (green), with either anti-G3BP1 or anti-p62 antibodies (red), together with Hoechst 33258 (blue). The proportions of cells containing TDP-43/G3BP1-positive SGs or TDP-43/p62-positive aggresomes are presented as the mean ± SD (n = 3). (E) iPSC-derived neurons were transfected with nontargeting siRNA (NT) or G3BP1-siRNA (G3BP1-1 or G3BP1-2) for 48 h, and whole-cell extracts were characterized by Western blotting using anti-G3BP1, anti-USP10, anti-β-actin, and anti-β-tubulin III (β-tub III) antibodies. (F) The proportions of cells containing TDP-43/PABP-positive SGs are presented as the mean ± SD (n = 3). Statistical significance was determined by a two-tailed unpaired t test. (G) The indicated cells were treated with 2 μM MG-132 for 4 h. Then, cells were stained with anti-TDP-43 antibodies (green), anti-PABP antibodies (red), and Hoechst 33258 (blue). (H) iPSC-derived neurons were transfected with nontargeting siRNA (NT) or USP10-siRNA (USP10-1 or USP10-2) for 48 h, and whole-cell extracts were characterized by Western blotting using anti-USP10, anti-G3BP1, anti-β-actin, and anti-β-tubulin III antibodies. (I) The proportions of cells containing TDP-43/p62-positive aggresomes are presented as the mean ± SD (n = 3). Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test. (J) The indicated cells were treated with 2 μM MG-132 for 8 h. Then, cells were stained with anti-TDP-43 antibodies (green), anti-p62 antibodies (red), and Hoechst 33258 (blue). The levels of statistical significance are as follows: *, P < 0.05; **, P < 0.01.

To examine how TDP-43 is localized in SGs and then in aggresomes, we characterized the localization of GFP-TDP-35 in SGs and aggresomes. GFP-TDP-35 overexpression in USP10-WT cells induced GFP-TDP-35-positive SGs (Fig. 7A). MG-132 or BTZ treatment for 12 h decreased the number of cells with these SGs but increased the number of cells with GFP-TDP-35-positive aggresomes (Fig. 7A and B). The previous study suggested that ubiquitinated proteins promote the clearance of SGs in cells under stress conditions (38). Thus, these results suggest that protein ubiquitination induced by PI treatment promotes the clearance of GFP-TDP-35-positive SGs and the formation of GFP-TDP-35-positive aggresomes. On the other hand, USP10-KD decreased the number of cells with BTZ-induced GFP-TDP-35-positive aggresomes and increased the number of cells with GFP-TDP-35-positive SGs (Fig. 7C). Furthermore, the overexpression of USP10 without PI treatment increased the number of cells with GFP-TDP-35-positive aggresomes to approximately 3% of the cells in the perinuclear region and decreased the number of cells with GFP-TDP-35-induced SGs (Fig. 7D and E). The decrease in the number of SG-positive cells by the overexpression of USP10 is consistent with a previous report that USP10 is a negative regulator of SG formation (22). Taken together, these results indicate that USP10 promotes the clearance of SGs and the formation of aggresomes.

FIG 7.

USP10 and p62 promote TDP-35-positive aggresome formation. (A) Neuro2a or 293T cells were transfected with the GFP-TDP-35 plasmid and were treated with 5 μM MG-132 or DMSO for 12 h. Then cells were stained with anti-HDAC6, anti-p62, or anti-USP10 antibodies (red) and Hoechst 33258 (blue). The scale bar indicates 10 μm (A, B, D, and I). (B) 293T cells were transfected with the GFP-TDP-35 plasmid, treated with 1 μM BTZ or DMSO for 12 h, and stained with anti-G3BP1 antibodies (red) or anti-HDAC6 antibodies (red) and Hoechst 33258 (blue). (C) Control 293T (NT) and USP10-KD (USP10-1) cells were transfected with the GFP-TDP-35 plasmid and treated with 1 μM BTZ or DMSO for 12 h. The cells were then stained with anti-G3BP1 or anti-HDAC6 antibodies together with Hoechst 33258. The proportions of cells containing GFP-TDP-35/G3BP1-positive SGs and GFP-TDP-35/HDAC6-positive aggresomes are presented as the mean ± SD (n = 3). Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test (C, F, and H). (D) 293T cells were transfected with the GFP-TDP-35 plasmid together with the HA-USP10 plasmid. The cells were then stained with anti-G3BP1 antibodies (red) or anti-HDAC6 antibodies (red) and Hoechst 33258 (blue). Simultaneously, whole-cell extracts prepared from the indicated 293T cells were characterized by Western blotting using anti-HA, anti-GFP, and anti-β-actin antibodies. (E) The proportions of cells containing GFP-TDP-35/G3BP1-positive SGs or GFP-TDP-35/HDAC6-positive aggresomes are presented as the mean ± SD (n = 3). Statistical significance was determined by a two-tailed unpaired t test. (F) 293T cells were transfected with the GFP-TDP-35 plasmid together with the FLAG-G3BP1 or/and HA-USP10 plasmid. The cells were then stained with anti-G3BP1 or anti-HDAC6 antibodies and Hoechst 33258. The proportions of cells containing GFP-TDP-35/G3BP1-positive SGs or GFP-TDP-35/HDAC6-positive aggresomes are presented as the mean ± SD (n = 3). (G) 293T cells were transfected with nontargeting siRNA (NT) or p62-siRNAs (p62-1, p62-2, or p62-3) for 48 h. Whole-cell extracts were characterized by Western blotting using anti-p62 and anti-α-tubulin antibodies. (H and I) 293T cells were transfected with nontargeting siRNA (NT) or p62-siRNAs (p62-2 or p62-3) and transfected with the GFP-TDP-35 plasmid. These cells were treated with 1 μM BTZ or DMSO for 12 h. The cells were then stained with anti-G3BP1 or anti-HDAC6 antibodies (red) and Hoechst 33258 (blue). The proportions of cells containing GFP-TDP-35/G3BP1-positive SGs or GFP-TDP35/HDAC6-positive aggresomes are presented as the mean ± SD (n = 3). (I) Immunofluorescence images of GFP-TDP-35/HDAC6 are shown. The levels of statistical significance are as follows: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

USP10 is a deubiquitinase for several substrates (39–42). However, a USP10 deubiquitinase-defective mutant (USP10^C424A) also decreased the number of cells with GFP-TDP-35-positive SGs and increased the number of cells with GFP-TDP-35-positive aggresomes to a level equivalent to that in USP10-WT cells, indicating that the deubiquitinase activity of USP10 is not required for the clearance of SGs and the formation of aggresomes (see Fig. S5). When USP10 was overexpressed, GFP without the TDP-35 fragment (GFP) also induced GFP/HDAC6-positive aggresomes; however, the frequency and size of aggresomes were decreased in comparison to GFP-TDP-35 (see Fig. S6). Collectively, these results suggested that USP10 promotes the clearance of SGs and the formation of aggresomes by a mechanism independent of deubiquitination.

As mentioned above (Fig. 3), G3BP1-KD decreased the number of cells with aggresomes in PI-treated cells. Therefore, we next examined whether G3BP1 affects the aggresome formation induced by USP10. Overexpression of G3BP1 increased the number of cells with aggresomes induced by the coexpression of USP10 and GFP-TDP-35 but not by GFP-TDP-35 alone (Fig. 7F; also see Fig. S7). These results suggested that G3BP1 augments GFP-TDP-35-positive aggresome formation induced by USP10 in 293T cells.

p62 promotes SG clearance and aggresome formation.

p62 is a Ub-binding protein that is required for the aggregation of many ubiquitinated proteins and promotes aggresome formation induced by USP10 (19). Furthermore, p62 has been shown to promote the clearance of SGs in cells exposed to heat shock (28). Thus, we examined whether p62 plays a role in the clearance of SGs in PI-treated cells. p62-KD increased the number of cells with GFP-TDP-35-positive SGs with or without PI treatment and decreased the number of cells with GFP-TDP-35-positive aggresome with PI treatment (Fig. 7G to I). These results suggested that p62 promotes the clearance of SGs and the formation of aggresomes in PI-treated cells.

The overexpression of p62 in cultured cells induces aggresome-like aggregates in the cytoplasm (19, 43). To further evaluate the function of p62 in the formation of TDP-43-positive aggresomes, we overexpressed p62 in 293T cells. Without the overexpression of p62, TDP-43 was mainly detected in the nucleus and was weakly detected in the cytoplasm of 293T cells. The overexpression of p62 induced small and large p62-positive aggregates in the cytoplasm, and large p62-positive aggregates were constantly colocalized with TDP-43 (see Fig. S8A). Large p62 aggregates were also colocalized with HDAC6 and USP10 (markers of aggresomes), suggesting that the large p62/TDP-43-positive aggregates were aggresomes (see Fig. S8B and C). The coexpression of p62 with GFP-TDP-35 induced GFP-TDP-35/p62-positive aggregates, and the number of GFP-TDP-35/p62-positive aggregates was greater than the number of GFP-TDP-43/p62-positive aggregates induced by the coexpression of p62 with GFP-TDP-43 (see Fig. S8D). When p62 was overexpressed, as in GFP-TDP-35 versus GFP-TDP-43, nontagged TDP-35 colocalized with p62-positive aggregates in the cytoplasm more frequently than did nontagged TDP-43 (see Fig. S8E). These results suggested that p62 promotes the formation of TDP-43/TDP-35-positive aggresomes in 293T cells and that TDP-35 localizes to aggresomes more efficiently than does TDP-43.

GFP-TDP-35 interacts with USP10 and G3BP1 through RNA.

To examine how USP10 promotes SG clearance and aggresome formation, we next examined whether GFP-TDP-35 and its mutants interact with USP10. An immunoprecipitation analysis detected the interaction of GFP-TDP-35 with USP10 (Fig. 8A and B). TDP-43 and TDP-35 interact with several RNA-binding proteins (44, 45). We examined whether TDP-35 interacts with USP10, G3BP1, and/or PABP through RNA. For this purpose, we used two RNA-binding-attenuated mutants of GFP-TDP-35, namely, GFP-TDP-35/5FL (46) and GFP-TDP-25 (Fig. 8A). GFP-TDP-35/5FL contains five amino acid substitutions from Phe (F) to Leu (L) in RRM1 and RRM2 of GFP-TDP-35 (47, 48). GFP-TDP-25 encodes amino acids 216-414 of TDP-43, and all of RRM1 and part of RRM2 of TDP-43 have been deleted. The loss of RRM1 of TDP-43 is reported to significantly reduce its RNA-binding activity (49). A coimmunoprecipitation assay showed that, while WT GFP-TDP-35 interacted with USP10, G3BP1, and PABP, RNA-binding-defective GFP-TDP-35/5FL and GFP-TDP-25 failed to interact with USP10, G3BP1, and PABP (Fig. 8B). The RNA-binding activity of TDP-43 is impaired by acetylation of TDP-43 at lysines 145 and 192 (50). Thus, we characterized acetylation-mimic (GFP-TDP-35-K145Q/K192Q [GFP-TDP-35/2KQ]) and acetylation-defective (GFP-TDP-35-K145R/K192R [GFP-TDP-35/2KR]) TDP-35 mutants. GFP-TDP-35/2KR, but not GFP-TDP-35/2KQ, interacted with USP10, G3BP1, and PABP (Fig. 8B). Furthermore, RNase treatment abolished the interaction of GFP-TDP-35 with USP10 and PABP (Fig. 8C). These results suggested that GFP-TDP-35 interacts with USP10, G3BP1, and PABP by interacting with RNA. We also examined the binding of endogenous TDP-43 with USP10, G3BP1, and PABP (Fig. 8D). An immunoprecipitation analysis showed that TDP-43 interacts with USP10, G3BP1, and PABP in 293T cells.

FIG 8.

TDP-35 interacts with G3BP1 and USP10 by associating with RNA. (A) A schematic illustration of the structure of the TDP-35 mutants used in this study. (B) 293T cells were transfected with the GFP-TDP-35 (TDP-35, TDP-35/5FL, TDP-35/2KQ, or TDP-35/2KR) plasmid or GFP-TDP-25 plasmid. Cell lysates prepared from transfected cells were immunoprecipitated with anti-GFP antibodies. The cell lysate (Input) and immunoprecipitates (IP) were characterized by Western blotting with anti-USP10, anti-G3BP1, anti-PABP, and anti-GFP antibodies. The asterisk indicates nonspecific bands. (C) 293T cells were transfected with the GFP-TDP-35 plasmid or the GFP-TDP-35/5FL plasmid. Then cells were lysed with ice-cold NP-40-lysis buffer containing RNase A, and the soluble cell lysates were immunoprecipitated with anti-GFP antibodies. The cell lysate (Input) and immunoprecipitate (IP) were subjected to Western blotting with anti-USP10, anti-PABP, or anti-GFP antibodies. (D) Cell lysates prepared from 293T cells were immunoprecipitated with anti-TDP-43 and control antibodies. The cell lysate (Input) and immunoprecipitates (IP) were characterized using Western blot analysis with anti-USP10, anti-G3BP1, anti-PABP, and anti-TDP-43 antibodies. (E and G) Neuro2a cells (E) or 293T cells (G) were transfected with the GFP-TDP-35 (TDP-35, TDP-35/5FL, TDP-35/2KQ, or TDP-35/2KR) or GFP-TDP-25 plasmid. The cells were fixed and stained with anti-USP10 antibodies (red) and Hoechst 33258 (blue). The scale bar indicates 10 μm (E and G). (F and H) GFP intensities in USP10-positive SGs in Neuro2a cells (F) or 293T cells (H) are presented as the mean ± SD (n = 30). Statistical significance was determined by a two-tailed unpaired t test or a one-way ANOVA followed by Tukey's multiple-comparison test. The levels of statistical significance are as follows: **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

RNA-binding-attenuated GFP-TDP-35 mutants show reduced localization in SGs and aggresomes.

We next examined whether the RNA-binding activity of GFP-TDP-35 is required for localization in SGs. The overexpression of three RNA-binding-defective GFP-TDP-35 mutants (GFP-TDP-35/5FL, GFP-TDP-35/2KQ, and GFP-TDP-25) induced USP10/GFP-TDP-positive SGs in Neuro2a cells (Fig. 8E and F) and 293T cells (Fig. 8G and H), but the amounts in SGs were less than those of two RNA-binding-active proteins (GFP-TDP-35-WT and GFP-TDP-35/2KR). These results suggested that the RNA-binding activity of GFP-TDP-35 is critical for efficient localization in SGs.

We next examined whether the RNA-binding activity of GFP-TDP-35 affects aggresome localization. Upon PI treatment, a portion of RNA-binding-defective GFP-TDP-35 mutants (GFP-TDP-35/5FL, GFP-TDP-35/2KQ, and GFP-TDP-25) were localized in USP10-positive aggresomes in 293T cells but there was less localization, in comparison to RNA-binding-competent proteins (GFP-TDP-35-WT and GFP-TDP-35/2KR); instead, they formed USP10-negative aggregates distinct from USP10-positive aggresomes (Fig. 9A and B). In addition, RNA-binding-defective GFP-TDP-35 mutants decreased the number of cells with USP10-induced aggresomes, relative to RNA-binding-competent GFP-TDP-35 (Fig. 9C and D). These results suggest that the RNA-binding activity of GFP-TDP-35 augments the localization of GFP-TDP-35 in aggresomes.

FIG 9.

RNA-binding activity of TDP-35 is required for efficient localization in SGs and aggresomes. (A) 293T cells were transfected with the GFP-TDP-35 (TDP-35, TDP-35/5FL, TDP-35/2KQ, or TDP-35/2KR) plasmid or the TDP-25 plasmid and further treated with 1 μM BTZ for 12 h. The cells were then fixed and stained with anti-USP10 antibodies (red) and Hoechst 33258 (blue). The scale bar indicates 10 μm (A, C, and E). Colocalization of GFP-TDP (green) and USP10 (red) was evaluated by a line profile analysis. (B) The proportions of USP10-positive/GFP-TDP-positive aggresomes (n = 3) and USP10-negative/GFP-TDP-positive aggregates (n = 3) are presented. Statistical significance was determined by a one-way ANOVA followed by Tukey's multiple-comparison test (B, D, and F). (C) 293T cells were transfected with the GFP-TDP-35 (TDP-35, TDP-35/5FL, TDP-35/2KQ, or TDP-35/2KR) plasmid or TDP-25 plasmid together with the HA-USP10 plasmid. The cells were fixed and stained with anti-HDAC6 or anti-HA antibodies (red) and Hoechst 33258 (blue). Colocalization of GFP-TDP (green) and USP10 (red) was evaluated by a line profile analysis. (D) The proportions of cells containing GFP-TDP/HDAC6-positive aggresomes are presented as the mean ± SD (n = 3). (E) 293T cells were transfected with the GFP-TDP-35 or GFP-TDP-35/5FL plasmid and treated with 0.5 mM sodium arsenite for 1 h. The cells were then fixed and stained with anti-G3BP1 antibodies (red) and Hoechst 33258 (blue). (F) The proportions of cells containing G3BP1-positive/GFP-TDP-35-positive SGs and G3BP1-negative/GFP-TDP-35-positive aggregates are presented as the mean ± SD (n = 3). The levels of statistical significance are as follows: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Sodium arsenite (an oxidant) is a well-known SG inducer in various types of cells (51). Cohen et al. showed that sodium arsenite treatment induces the acetylation of TDP-43 at lysines 145 and 192 in cultured cells and the acetylation of TDP-43 reduces the RNA-binding activity (50). Thus, we next examined whether arsenite treatment induces G3BP1-negative GFP-TDP-35 aggregation by reducing the RNA-binding activity of GFP-TDP-35. The overexpression of GFP-TDP-35 increased the number of cells with GFP-TDP-35/G3BP1-positive SGs in 293T cells; however, the number was decreased by arsenite treatment. Instead, arsenite treatment generated G3BP1-negative GFP-TDP-35 aggregates resembling those induced by RNA-binding-defective GFP-TDP-35 mutants in PI-treated cells (Fig. 9E and F). In addition, arsenite treatment increased the number of cells with G3BP1-negative GFP-TDP-35/5FL aggregates (Fig. 9E and F). Taken together, these results suggested that arsenite treatment reduces the localization of GFP-TDP-35 in SGs and induces G3BP1-negative GFP-TDP-35 aggregates by reducing the RNA-binding activity of GFP-TDP-35. It should be noted that arsenite treatment increased the number of cells with G3BP1-negative GFP-TDP-35/5FL aggregates (Fig. 9F). It has been reported that the RNA-binding activity of TDP-43/5FL is strongly impaired but not completely lost (46, 48). Therefore, these results suggest that arsenite treatment further reduces the RNA-binding activity of GFP-TDP-35/5FL, thereby inducing an increase in G3BP1-negative GFP-TDP-35/5FL aggregates.

TDP-43 forms USP10-negative aggregates in the cytoplasm of ALS neurons.

Cytoplasmic aggregation of phosphorylated TDP-43 (pTDP-43) is a hallmark pathology of ALS neurons (52, 53). We examined whether pTDP-43 aggregates in ALS neurons from seven ALS cases were colocalized with USP10 by immunohistochemistry. Cytoplasmic pTDP-43 aggregates were detected only in the neurons of ALS patients and not in controls, and most of the cytoplasmic pTDP-43 aggregates in ALS neurons were not colocalized with USP10 (Fig. 10A to C). Elderly brain specimens may have autofluorescent lipofuscin granules. However, USP10-negative pTDP-43 aggregates were detected in samples in which such autofluorescence was reduced by an autofluorescence elimination reagent (see Fig. S9A). These results suggested that pTDP-43 forms USP10-negative aggregates in ALS neurons. Some pTDP-43 aggregates in ALS neurons showed colocalization with USP10, although the number was less than that of USP10-negaitve pTDP-43 aggregates (Fig. 10B and C; also see Fig. S9B and C).

FIG 10.

Aberrant TDP-43 aggregates in ALS neurons do not colocalize with USP10. (A) Representative double immunostaining of pTDP-43 and USP10 in ALS patients and a control subject. The colocalization of pTDP-43 (green) and USP10 (red) was evaluated by a line profile analysis. The line profile analysis was done along the dotted line of the arrow. Asterisks indicate autofluorescent lipofuscin granules. The scale bar indicates 20 μm. (B) The proportions of cells containing USP10-negative or USP10-positive pTDP-43 aggregates are presented as the mean ± SD (n = 7). Statistical significance was determined by a two-tailed unpaired t test. (C) The numbers of cells containing USP10-negative or USP10-positive pTDP-43 aggregates from seven ALS cases are shown. The level of statistical significance is as follows: ****, P < 0.0001.

DISCUSSION

The aggregation of ubiquitinated insoluble TDP-43/TDP-35 in the cytoplasm of neurons is a hallmark pathology of ALS and FTLD-TDP (3, 4). TDP-43 localizes to SGs and aggresomes under various stress conditions (31–33). Accumulating evidence suggests that aberrant SGs with delayed clearance and a large amount of ubiquitinated proteins induce pathological aggregation of various RNA-binding proteins, including TDP-43 (9, 24, 26). In this study, we found that USP10 promotes the clearance of TDP-43/TDP-35-positive SGs and the formation of TDP-43/TDP-35-positive aggresomes in PI-treated cells. Depletion of USP10 delayed SG clearance, increased the amounts of Ub in SGs (Fig. 5), and increased the amount of insoluble TDP-35 in the cytoplasm (Fig. 4). These results suggest that USP10 promotes SG clearance and aggresome formation in cells under stress conditions, and the dysfunction of USP10 may promote pathological aggregation of insoluble TDP-43/TDP-35 in the cytoplasm (see Fig. S9D in the supplemental material).

Like USP10, p62 promoted the clearance of SGs in PI-treated cells (Fig. 7H). This is consistent with the report that p62 promotes the clearance of SGs in heat shock-treated cells (28). p62 has been shown to interact with USP10 (19, 23). Thus, these results suggest that p62 and USP10 interact to cooperatively promote SG clearance and aggresome formation in stressed cells. In addition, the p62 gene is a causative factor of both ALS and FTLD (54–56). These results suggest that defective regulation of SGs and aggresomes by p62 mutants may be involved in the development of ALS in some ALS patients with p62 mutations.

The present study suggested that ubiquitinated proteins promote SG clearance. First, PI treatment, which increases the amount of ubiquitinated proteins, promoted the clearance of GFP-TDP-35-positive SGs and simultaneously promoted the formation of GFP-TDP-35-positive aggresomes (Fig. 7A and B). Next, depletion of p62, a Ub-binding protein, delayed the clearance of SGs (Fig. 7H). These results suggest that ubiquitinated proteins accumulated in PI-treated cells interact with and activate p62 and this activated p62, together with USP10, initiates SG clearance and aggresome formation (see Fig. S9D).

Similar to TDP-43, several pathological RNA-binding proteins, such as FUS, localize to both SGs and aggresomes under stress conditions (26, 57). The present study presented the possibility that not only TDP-43 but also other pathological RNA-binding proteins in SGs are transported from SGs to aggresomes and the transport is facilitated by G3BP1 and USP10. Depletion of either G3BP1 or USP10 reduced the formation of aggresomes induced by PI treatment (Fig. 3 and 4). Furthermore, GFP-TDP-35 was transported from SGs to aggresomes by either the overexpression of USP10 or PI treatment (Fig. 7). G3BP1 and USP10 interact with several RNA-binding proteins present in SGs (13). Therefore, these results suggest that G3BP1 and USP10 promote the transport of pathological RNA-binding proteins from SGs to aggresomes and inhibit the aberrant aggregations of these pathological RNA-binding proteins.

In contrast to the present findings, G3BP1 has been shown to inhibit the formation of aggresomes in cells overexpressing ubiquitination/aggregation-prone proteins (23). CFTRΔF508 is ubiquitination/aggregation prone and is a causative factor of familial cystic fibrosis, and its overexpression in cells induces the formation of CFTRΔF508-positive aggresomes (58). It has been reported that depletion of G3BP1 increases the formation of CFTRΔF508-positive aggresomes induced by overexpression of CFTRΔF508 and this increase is eliminated by the depletion of either USP10 or p62 (23). Taken together, the present and previous findings suggest that G3BP1 has both inhibitory and stimulatory effects on aggresome formation. Treatment of cells with PI induced the formation of SGs first, followed by aggresomes (Fig. 1B and D). Therefore, these results suggest that G3BP1 inhibits the formation of aggresomes during the stress-induced SG formation phase and promotes the formation of aggresomes during the SG clearance phase.

We also found that USP10 and G3BP1 interact with TDP-43/TDP-35 in an RNA-dependent manner in cultured cells and RNA-binding-defective TDP-35 mutants do not interact with G3BP1 and USP10, inducing aberrant G3BP1/USP10-negative aggregates distinct from SGs and aggresomes (Fig. 8 and 9). Furthermore, some pathological cytoplasmic TDP-43/TDP-35 aggregates in the neurons of ALS patients were USP10 negative (Fig. 10). These results suggested that RNA-binding-defective TDP-43/TDP-35 mutants form pathological TDP-43/TDP-35 aggregates in ALS neurons (see Fig. S9D). Cohen et al. showed that ALS neurons have RNA-binding-defective acetylated TDP-43 in their cytoplasm (50). Although some RNA-binding proteins (fragile X mental retardation protein [FMRP], TIA1, and TIAR) are colocalized with TDP-43 in SGs in cultured cells, these RNA-binding proteins were not colocalized with TDP-43 aggregates in the cytoplasm of ALS neurons (59). Taken together, our results suggest that reduced RNA binding of TDP-43/TDP-35 impairs the localization of TDP-43/TDP-35 in SGs and aggresomes and promotes pathological TDP-43/TDP-35 aggregates in ALS neurons.

Cohen et al. showed that the RNA-binding activity of TDP-43 is reduced by the stress-induced acetylation of TDP-43 at lysines 145 and 192 and the acetylation induces insoluble aberrant TDP-43 aggregates in the cytoplasm (50). We also observed that an acetylation-mimic TDP-35 mutant produced USP10-negative GFP-TDP-35 aggregates distinct from SGs and aggresomes (Fig. 9A and B). Importantly, an acetylated TDP-43 at K145 was detected in pTDP-43-positive inclusions of spinal cord lesions of all six ALS patients who were tested but not in patients with FTLD-TDP (50). Taken together, these results suggest that acetylation-induced reduced RNA binding of TDP-43 is a key mechanism for the generation of TDP-43/TDP-35 aggregates in ALS patients but not in patients with FTLD-TDP.

Like other RNA-binding-defective GFP-TDP-35 mutants, GFP-TDP-25 generated USP10-negative aberrant cytoplasmic aggregates in PI-treated cells (Fig. 9A and B). Several studies suggested that TDP-25 plays a causative role in the development of ALS and/or FTLD-TDP. For instance, TDP-25 is frequently detected in neuronal cytoplasmic inclusions in ALS and FTLD-TDP patients (60). Moreover, animals carrying the TDP-25 transgene develop ALS/FTLD-like phenotypes (61, 62). These results suggest that the cleavage-induced generation of RNA-binding-defective TDP-25 protein is a factor initiating aberrant pathological aggregation in ALS and/or FTLD-TDP patients. It should be noted that TDP-25 does not have the two acetylation sites of TDP-43 at K145 and K192, due to N-terminal deletion (Fig. 8A). Thus, acetylation-negative TDP-43 inclusions detected in FTLD/TDP neurons in the study by Cohen et al. (50) might be explained by the accumulation of TDP-25 aggregates in FTLD/TDP neurons.

TDP-35 is induced by a protease-dependent cleavage of TDP-43, and its production is increased by various stress stimuli (29). The present study proposes the intriguing hypothesis that TDP-35, but not TDP-43, is responsible for the initial formation of pathological TDP-43/TDP-35 aggregates in ALS. TDP-35 localized to SGs much more efficiently than TDP-43 (see Fig. S3A and B). Decreasing USP10 delayed SG clearance in stressed cells and simultaneously increased the amount of cytoplasmic insoluble TDP-35 but not TDP-43 (Fig. 4 and 5). Once TDP-35 forms cytoplasmic aggregates, these TDP-35 aggregates facilitate the translocation of TDP-43 from the nucleus to the cytoplasm, recruiting and incorporating TDP-43 into the aggregates and leading to the formation of TDP-43/TDP-35 aggregates that are detected in ALS neurons (44). A further analysis is needed to understand how TDP-43 and/or TDP-35 form pathological aggregates in the first step of ALS development.

MATERIALS AND METHODS

Cell lines and culture conditions.

Neuro2a is a mouse neuroblastoma cell line. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 5% heat-inactivated fetal bovine serum (FBS), Opti-MEM (Thermo Fisher Scientific), antibiotics (100 units/ml penicillin and 100 μg/ml streptomycin), and minimal essential medium nonessential amino acid solution (MEM NEAA) (Thermo Fisher Scientific). 293T cells were cultured in DMEM supplemented with 10% FBS, 4 mM l-glutamine, antibiotics, and MEM NEAA.

Preparation of iPSC-derived neurons.

Human iPSCs (clone1231A3) (63) were obtained from Riken BioResource Research Center (BRC). iPSCs were cultured according to the feeder-free culture method established by the Center for iPS Cell Research and Application (CiRA), Kyoto University. The iPSCs were suspended in iPSC medium (StemFit AK02N; AJINOMOTO) and cultured in 6-well culture plates coated with laminin-511 E8 protein (laminin-positive 6-well plates). After at least one passage, cells were cultured in laminin-positive 6-well plates at 1 × 10⁵ cells/well. The next day, the cells were cultured in pluripotent stem cell (PSC) neural induction medium (Thermo Fisher Scientific) for 1 week to induce differentiation into neural stem cells (NSCs). Then, cells were suspended in StemPro NSC serum-free medium and cultured in Matrigel-coated 6-well plates at 2 × 10⁵ cells/well (Thermo Fisher Scientific). On the second day after culture, the cells were cultured in 2% B27/neurobasal medium to induce differentiation into neurons. On day 9 after culture, cells were cultured in Matrigel-coated coverslips in 6-well plates. Two days later, the cells were used for experiments.

Preparation of primary rat neuron-enriched cells.

Sprague-Dawley rats (Japan SLC, Inc., Shizuoka, Japan) were bred in the animal care facility of Niigata University Brain Research Institute. The rat experiment was approved by the Animal Use and Care Committee of Niigata University and was carried out in accordance with the institutional guidelines and the guidelines of the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals (NIH publications 80-23). Rat primary neuron-enriched cells were prepared as described previously (64). Briefly, cortical tissues were dissected from rat embryos at embryonic days 16 to 19 and digested with papain (Worthington Biochemical Corp.). Cells prepared from cortical tissues were seeded on glass coverslips coated with poly-d-lysine (Sigma-Aldrich) in a 6-well plate (Corning) and cultured in DMEM supplemented with 10% FBS. After overnight culturing, the medium was replaced with serum-free DMEM supplemented with N-2 supplement (Gibco) and l-glutamine, and the cells were further cultured for 3 days. Cells were then treated with 2 μM MG-132 for 0 to 8 h, and formation of SGs and aggresome-like aggregates in cells was characterized by immunostaining with anti-TDP-43 antibodies (green) together with either anti-G3BP1 antibodies (red) or anti-p62 antibodies (red), respectively.

Reagents and antibodies.

The following reagents were purchased from the indicated companies: MG-132 (474790; Calbiochem), BTZ (B-1408; LC Laboratories), sodium arsenite (Wako Pure Chemical Industries), VER (sc-358808; Santa Cruz Biotechnology), puromycin (P8833; Sigma-Aldrich), and Hoechst 33258 (H3569; Molecular Probes). The following antibodies were used in this study: anti-USP10 (A300-901A; Bethyl Laboratories; or HPA006731; Sigma-Aldrich), anti-TDP-43 (12892-1-AP; Proteintech), anti-pTDP-43 (TIP-PTD-M01; Cosmo Bio), anti-Ub (sc-8017; Santa Cruz Biotechnology; or ab7780; abcam), anti-p62 (PM045; MBL; or GP62-C; PROGEN), anti-HDAC6 (sc-11420; Santa Cruz Biotechnology), anti-G3BP (611127; BD Transduction Laboratories), anti-G3BP2 (A302-040A; Bethyl Laboratories), anti-PABP (ab21060; Abcam; or c-18611; Santa Cruz Biotechnology), anti-hemagglutinin (HA) (H9658; Sigma-Aldrich; or 3724; Cell Signaling), anti-GFP (sc-9996; Santa Cruz Biotechnology), anti-HDAC1 (05-614; Upstate), anti-α-tubulin (DM1A; Calbiochem), anti-lamin B1 (sc-374015; Santa Cruz Biotechnology), anti-β-tubulin III (G7121; Promega), anti-MAP-2 (MAB3418; Merck Millipore), and anti-β-actin (sc-47778; Santa Cruz Biotechnology).

Plasmids.

Expression plasmids encoding WT USP10 (HA-USP10) with an N-terminal HA epitope tag or deubiquitinase-defective HA-USP10^C424A were generated by subcloning the respective USP10 cDNAs into an expression plasmid (pCMV-HA; Clontech). An expression plasmid encoding FLAG-tagged G3BP1 (FLAG-G3BP1) was constructed by inserting human G3BP1 cDNA into pFLAG-CMV2 vector (Sigma-Aldrich) (13). Expression plasmids encoding TDP-43 or TDP-35 were generated by subcloning the respective cDNAs into an expression plasmid (pcDNA3; Invitrogen). Expression plasmids encoding TDP-43, TDP-35, or TDP-25 with an N-terminal GFP tag were generated by subcloning the respective cDNAs into an expression plasmid (pEGFP-C3; Clontech). GFP-TDP-35/5FL has five mutations on RMMs (47) and was inserted into pEGFP-C3. Mutations of TDP-35 from Lys (K) to Gln (Q) at residues K145 and K192 were introduced by PCR-based mutagenesis and inserted into pEGFP-C3 to construct GFP-TDP-35/2KQ. GFP-TDP-35/2KR has mutations from K to Arg (R) at residues K145 and K192. A retroviral expression plasmid (pMXs-puro) encoding p62 was used for the expression of exogenous p62. The lentiviral short hairpin RNA (shRNA) plasmid (pLKO.1-puro) targeting human USP10 mRNA with a puromycin resistance gene was purchased from Sigma-Aldrich.

Plasmid transfection.

Plasmids (0.5 to 1.0 μg) were transfected into Neuro2a cells (1.0 × 10⁵ cells) or 293T cells (1.5 × 10⁵ cells) on a 6-well plate (Corning) by using the FuGENE 6 reagent according to the manufacturer’s instructions (Promega).

Establishment of stable cell lines by viral transduction.

Stable KDs of endogenous USP10 in 293T cells were carried out by using the lentivirus vector encoding USP10-shRNA. HIV-1-based lentiviruses encoding USP10-shRNA were produced by cotransfection of the three plasmids (pLKO.1-puro [4.28 μg], pCAG-HIVgp [2.86 μg], and pCMV-VSV-G-RSV-Rev [2.86 μg]) into 293T cells (2 × 10⁶ cells) using FuGENE HD reagent according to the manufacturer’s instructions (Promega), the viruses were concentrated with Amicon Ultra-15 units (Millipore) to increase the infectious titer of the virus, and 293T cells were infected in the presence of 8 μg/μl Polybrene. These cells were cultured in selection medium containing 2 μg/ml puromycin.

RNA interference.

siRNAs specific to human G3BP2 RNA (oligonucleotide identification number HSS114988), human USP10 RNA (oligonucleotide identification numbers HSS113446 [USP10-1] and HSS113447 [USP10-2]), and human p62 RNA (oligonucleotide identification numbers HSS113118 [p62-1], HSS113117 [p62-2], and HSS113116 [p62-3]) and the negative-control siRNA (catalogue number 12935-100) were purchased from Invitrogen. FlexiTube siRNAs specific to G3BP1 (G3BP1-1) (catalogue number S102665194) and Allstars negative-control siRNA (catalogue number 1027280) were purchased from Qiagen. MISSION siRNA specific to the 3′ untranslated region of G3BP1 (G3BP1-2) (siRNA identification number SASI_Hs01_00045804) and MISSION universal negative-control siRNA (SIC001) were purchased from Sigma-Aldrich. These siRNAs (20 to 100 pmol) were transfected into cells using Lipofectamine RNAiMAX reagents according to the manufacturer’s protocol (Invitrogen).

Coimmunoprecipitation assay.

293T cells (1.0 × 10⁷ cells) on a 10-cm² dish (Corning) were lysed with ice-cold NP-40 lysis buffer (1% Nonidet P-40, 25 mM Tris-HCl [pH 7.2], 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 20 μg/ml aprotinin), and cell lysates were treated with anti-GFP or anti-TDP-43 antibodies. Immune complexes were precipitated with protein G-Sepharose beads (GE Healthcare). The beads were washed and boiled in SDS lysis buffer (62.5 mM Tris-HCl [pH 6.8], 2% SDS, 10% glycerol, 5% 2-mercaptoethanol, and 0.005% bromophenol blue), and then the proteins released from the beads were subjected to Western blotting. To assess the RNA-dependent interaction of GFP-TDP-35 with USP10 or PABP, RNase A (313-01461; Nippongene), at a final concentration of 100 μg/ml, was added to the lysis buffer, the binding buffer, and the washing buffer in an immunoprecipitation assay.

Western blotting.

293T cells or iPSC-derived neurons were lysed with SDS lysis buffer, and cell lysates (10 to 20 μg of proteins) were separated by SDS-PAGE and electrophoretically transferred onto a polyvinylidene difluoride (PVDF) membrane (Immobilon; Millipore). The membrane was incubated with phosphate-buffered saline (PBS) containing 5% skim milk to block nonspecific antibody binding and then was treated with the indicated antibodies diluted in Can Get Signal solution (TOYOBO). Immunoreactive bands were detected with an enhanced chemiluminescence (ECL) detection system (ECL Western blotting detection reagents [GE Healthcare] or Pierce ECL Plus Western blotting substrate [Thermo Fisher Scientific]) and were visualized with Amersham Hyperfilm ECL films (Amersham). To measure the amounts of TDP-43 and TDP-35, the intensities of bands in Western blotting were measured using the ImageJ software program (NIH).

Preparation of cytoplasmic and nuclear extracts.

Cytoplasmic and nuclear extracts were isolated using the NE-PER nuclear and cytoplasmic extraction reagents (Thermo Fisher Scientific), and extracts were subjected to Western blotting.

Immunofluorescence analysis.

293T cells, Neuro2a cells, or iPSC-derived neurons were cultured on glass coverslips, and cells were fixed with 3.7% formaldehyde in PBS and permeabilized with 0.1% Triton X-100 in PBS. Cells were then treated with the primary antibodies and further incubated with the secondary antibody. The secondary antibodies used were anti-mouse IgG, anti-rabbit IgG, or anti-guinea pig IgG conjugated with Alexa Fluor 488 or Alexa Fluor 594 (Molecular Probes). Cell nuclei were stained with Hoechst 33258. The samples were mounted in Fluoromount/Plus (Diagnostic Biosystems), and the images were analyzed with a fluorescence microscope (BZ-8000 or BZ-X810; Keyence).

Microscopic analysis.

To measure SG formation, more than 300 cells in random fields from three coverslips were analyzed by staining with SG markers (anti-G3BP1, anti-PABP, or anti-USP10 antibodies). The percentages of cells with SGs were calculated as the ratio of cells with SGs positive for G3BP1, PABP, or USP10 to total cells. To measure cells with an aggresome, more than 300 cells in random fields from three coverslips were analyzed by staining with anti-HDAC6 or anti-p62 antibodies. The percentages of cells with an aggresome (more than 15 μm² in size) in the perinuclear regions with nuclear deformity were calculated as the ratio of aggresome-positive cells to total cells. To measure the size of GFP-TDP-35-positive SGs, the sizes of SGs were quantified from 30 randomly selected GFP-TDP-35-positive SGs using a fluorescence microscope with the ImageJ software program (NIH). To measure the localization of GFP-TDP-35/TDP-25 in SGs, the GFP fluorescence intensity of GFP-TDP-35/TDP-25 in SGs was measured from 30 randomly selected USP10-positive SGs using a BZ-II analyzer (Keyence). To measure cells with USP10- and GFP-TDP-35/TDP-25-positive aggresomes or USP10-negative and GFP-TDP-35/TDP-25-positive aggregates, more than 300 cells in random fields from three coverslips were analyzed by anti-USP10 antibody staining. Percentages of cells with USP10- and TDP-positive aggresomes or USP10-negative and TDP-positive aggregates were calculated as the ratio of cells with either USP10- and TDP-positive aggresomes or USP10-negative and TDP-positive aggregates to total GFP-positive cells. To measure the size of aggresomes, the sizes of aggresomes were quantified from 30 randomly selected aggresomes using a fluorescence microscope with the ImageJ software program (NIH).

Colocalization assay.

The colocalization of USP10 and exogenous GFP-TDP-35/TDP-25 or endogenous pTDP-43 was examined by a line profile analysis. The staining levels on the indicated lines in images were measured by the ImageJ software program (NIH).

Detection of p62-positive aggregates.

To measure cells with aggregates positive for p62/TDP-43, p62/HDAC6, or p62/USP10, 293T cells on coverslips were transfected with the p62 plasmid, and cells were stained with anti-p62 antibodies (red) together either with anti-TDP-43, anti-HDAC6, or anti-USP10 antibodies (green). More than 300 cells in random fields from three coverslips were analyzed to calculate the ratio of cells with either p62/TDP-43, p62/HDAC6, or p62/USP10 aggregates to cells with p62 aggregates. The colocalization of p62 and exogenous GFP-TDP-43/TDP-35 or endogenous TDP-43 was examined by a line profile analysis.

Pathological analyses.

The study using human brains was performed with the approval of the ethics committees of Niigata University (approval number 2018-0346). Spinal cords (C7, Th8, and L4) of ALS patients (n = 7) and control subjects (n = 3) were used. Formalin-fixed, paraffin-embedded, 4-μm-thick sections that had been pretreated by heating were immunostained with a rabbit anti-USP10 polyclonal antibody (HPA006731; Sigma-Aldrich) diluted 1:2,000 and a mouse anti-pTDP-43 monoclonal antibody (TIP-PTD-M01; Cosmo Bio) diluted 1:4,000 in PBS containing 1% bovine serum albumin (BSA). Antibody binding was detected by anti-rabbit immunoglobulin conjugated with Alexa Fluor 568 and anti-mouse immunoglobulin conjugated with Alexa Fluor 488 (Molecular Probes). To reduce lipofuscin-induced autofluorescence, brain specimens (n = 3) were treated with autofluorescence elimination reagent (Merck Millipore) after immunofluorescence staining and visualized using Alexa Fluor 647, with fluorescence wavelengths longer than the autofluorescence wavelength range. Immunofluorescence images were captured with a confocal microscope (FV3000RS; Olympus, Tokyo, Japan).

Statistical analysis.

Data were analyzed using a two-tailed unpaired t test or a one-way analysis of variance (ANOVA) followed by Tukey's multiple-comparison test using the Prism7 software program (GraphPad) and are presented as the mean ± standard deviation (SD).

Data availability.

The data that support the findings of this study are available from the corresponding author upon reasonable request.