Suppression of Linear Ubiquitination Ameliorates Cytoplasmic Aggregation of Truncated TDP-43

TAR DNA-binding protein 43 (TDP-43) is a predominant component of inclusions in the brains and spines of patients with amyotrophic lateral sclerosis (ALS). The progressive accumulation of inclusions leads to proteinopathy in neurons. We have previously shown that Met1(M1)-linked linear ubiquitin, which is specifically generated by the linear ubiquitin chain assembly complex (LUBAC), is colocalized with TDP-43 inclusions in neurons from optineurin-associated familial and sporadic ALS patients, and affects NF-κB activation and apoptosis. To examine the effects of LUBAC-mediated linear ubiquitination on TDP-43 proteinopathies, we performed cell biological analyses using full-length and truncated forms of the ALS-associated Ala315→Thr (A315T) mutant of TDP-43 in Neuro2a cells. The truncated A315T mutants of TDP-43, which lack a nuclear localization signal, efficiently generated cytoplasmic aggregates that were colocalized with multiple ubiquitin chains such as M1-, Lys(K)48-, and K63-chains. Genetic ablation of HOIP or treatment with a LUBAC inhibitor, HOIPIN-8, suppressed the cytoplasmic aggregation of A315T mutants of TDP-43. Moreover, the enhanced TNF-α-mediated NF-κB activity by truncated TDP-43 mutants was eliminated in the presence of HOIPIN-8. These results suggest that multiple ubiquitinations of TDP-43 including M1-ubiquitin affect protein aggregation and inflammatory responses in vitro, and therefore, LUBAC inhibition ameliorates TDP-43 proteinopathy.


Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the progressive degeneration of motor neurons in the brain and spinal cord [1,2]. Although most ALS cases are sporadic (sALS),~10% of ALS patients are familial (fALS) and over 30 potential ALS genes have been identified in fALS patients. These gene products seem to regulate multiple cellular functions such as RNA metabolism and proteostasis, protein trafficking, ubiquitin-proteasomal, and autophagic degradation, cytoskeletal and axonal dynamics, and neuroinflammation. TARDBP encodes TAR DNA-binding protein 43 , which is a key component of the insoluble and ubiquitinated cytoplasmic inclusions in the brain and spinal cord of ALS and frontotemporal dementia (FTD) patients [3]. TDP-43 contains two RNA recognition motifs (RRM1 and RRM2) at the N-terminal region,

Plasmids
The plasmids containing cDNAs of human wild type (WT) and the ALS-related Ala315→Thr (A315T) mutant of TDP-43 were kind gifts from Prof. Yoshitaka Nagai (Kindai University, Osaka-Sayama, Japan). The coding sequences were tagged with GFP, amplified by PCR, and cloned into the pcDNA3.1 expression vector (Invitrogen, Carlsbad, CA, USA). The open reading frames of the mouse and human LUBAC subunit cDNAs were amplified by reverse-transcription PCR. The truncation and deletion mutants of TDP-43 cDNAs were prepared by PCR, and the nucleotide sequences were verified. The cDNAs were ligated to the appropriate epitope sequences and cloned into the pcDNA3.1 or pCSII-EF-IRES2 vector (RIKEN BioResource Research Center, Tsukuba, Japan).

Immunofluorescent Staining and Imaging Analyses
Neuro2a cells (4 × 10 4 cells) were seeded on poly-D-lysine-coated coverslips in 24-well plates, one day before transfection. The plasmids encoding TDP-43 and/or LUBAC were transfected into Neuro2a cells, using PEI. After 24 h, the cells were fixed with phosphate buffered 4% paraformaldehyde at room temperature for 15 min, and then permeabilized/blocked in staining buffer (0.05 % saponin, 10 % FBS, 10 mM glycine in PBS) for 30 min. The cells were incubated overnight with primary antibodies at 4 • C in a humidity box. The next day, the cells were washed, incubated with secondary antibodies and DAPI, and then mounted onto glass slides with FluorSave (MilliporeSigma, Burlington, MA, USA). The confocal fluorescence images of the prepared slides were captured with an LSM700 or LSM800 confocal microscope system (Carl Zeiss, Oberkochen, Germany). All images were acquired as 16-bit depth images with a 20× dry, 40× oil-, or 63× water-immersion objective lens by scanning each channel two separate times for averaging. In quantitative imaging analyses using the Tet-On inducible TDP-43 expressing cells, fluorescence images of TDP-43-expressing cells in 24-well plates were randomly acquired by an IN Cell Analyzer 2500HS with a 60× objective lens (GE Healthcare, Chicago, IL, USA). Fluorescence intensities and sizes of TDP-43 aggregates in the images were analyzed by Cell Profiler ver. 4.2.1 (Broad Institute Inc., Cambridge, MA, USA) [27].

Construction of Hoip-Knockout Neuro2a Cells
The 5 -AGGAGCTGGCGAGCGCCCTGAGG-3 nucleotide sequences in exon 1 of the mouse Hoip gene were selected as the targets. Double-stranded target DNAs were prepared by annealing synthesized oligo DNAs, and then incorporated into the BbsI site of the px458 CRISPR/Cas9 vector (Addgene, Watertown, MA, USA). Each plasmid was transiently introduced into Neuro2a cells, and 2 days after transfection, the GFP positive cells were sorted by a FACSAria IIIu (BD Biosciences, San Jose, CA, USA). Single clones were obtained by limiting dilution. Genome editing of the Hoip gene was confirmed by genome sequencing and immunoblotting.

Statistics
The Mann-Whitney U test, Kruskal-Wallis test, t-test, and one-way ANOVA followed by a post hoc Tukey HSD test were performed using the GraphPad Prism 8 software. For all tests, a p value of less than 0.05 was considered statistically significant.

Expression of TDP-43 Truncations Generates Multiple Ubiquitin Chain-Positive Cytoplasmic Aggregates
To investigate the involvement of M1-linked linear ubiquitination in TDP-43 proteinopathy, we first constructed GFP-tagged full-length and truncated forms of the ALSassociated A315T mutant of TDP-43 ( Figure 1A), and ectopically expressed them in mouse neuroblastoma Neuro2a cells. Immunofluorescence analyses demonstrated that the GFP alone was expressed throughout the cell, and the A315T 1-414 -GFP was localized predominantly in the nucleus without aggregation. In contrast, the truncated mutants of TDP-43 such as A315T 89-414 -GFP, A315T 247-414 -GFP, and A315T 1-414 -∆NLS-GFP efficiently formed cytoplasmic aggregates, which were colocalized with the pan-and M1-linked ubiquitins ( Figures 1B and S1), suggesting that TDP-43 mutants lacking the NLS generated linear ubiquitin-positive cytoplasmic aggregates in vitro.
To further investigate the linkages of the ubiquitin chains in the cytoplasmic TDP-43 aggregates, we next examined the colocalization of M1-and proteasomal degradationinducible K48-linked ubiquitin chains (Figures 2A and S2). The M1-ubiquitin-positive cytoplasmic aggregates of A315T 89-414 -GFP and A315T 247-414 -GFP, but not GFP and A315T 1-414 -GFP, also reacted with the anti-K48-ubiquitin antibody, suggesting that the cytoplasmic aggregates of truncated TDP-43 include both M1-and K48-ubiquitin chains. In the presence of the proteasomal inhibitor, MG-132, the formation of cytoplasmic aggregates of TDP-35 and TDP-25 was reportedly upregulated and stabilized [7]. Therefore, to examine the effects of inhibitors of proteasomal or lysosomal/autophagic degradation on TDP-43 aggregate formation, we treated A315T-GFPs-expressing Neuro2a cells with MG-132 or chloroquine. The cells were lysed with RIPA buffer to obtain the detergent-soluble fraction, and the resultant pellet was further solubilized with 6 M urea-containing buffer to obtain the RIPAinsoluble/urea-soluble aggregate fraction. As shown in Figure 2B, the number of truncated forms of A315T-GFPs in the insoluble aggregate fraction were increased in the presence of MG-132, but not chloroquine. Moreover, A315T 247-414 -GFP migrated as lower molecular weight bands in both the RIPA-soluble and urea-soluble fractions in the absence or presence of MG-132 and chloroquine, indicating its susceptibility to endogenous proteolytic enzymes. A smeared band, representing the putative polyubiquitination of A315T 247-414 -GFP, was detected in the RIPA-insoluble and urea-soluble fraction. These results suggest that the inhibition of proteasome activity facilitates the aggregation of A315T-GFPs. In MG-132-treated Neuro2a cells, the cytoplasmic aggregates of A315T 89-414 -GFP and A315T 247-414 -GFP were increased, and the colocalization of M1-and pan-ubiquitins ( Figure 2C) as well as the M1and K48-ubiquitins ( Figure 2D) with the A315T-GFP truncated aggregates was detected.  Similarly, the colocalization of the K63-and M1-ubiquitins was detected in the cytoplasmic aggregates of A315T 89-414 -GFP and A315T 247-414 -GFP, but not GFP and A315T 1-414 -GFP ( Figure 3A). In the presence of MG-132, the cytoplasmic aggregates of A315T 89-414 -GFP and A315T 247-414 -GFP colocalized with K63-and M1-ubiquitins were increased ( Figure 3B). Collectively, these results suggest that multiple ubiquitin chains such as M1-, K48-, and K63-ubiquitin chains are cooperatively conjugated to the cytoplasmic aggregates of truncated TDP-43 mutants, and the inhibition of proteasome activity, which basically affects the K48-linked ubiquitin chain, increases the number of cytoplasmic aggregates containing multiple ubiquitin chains.

LUBAC Plays an Important Role in Cytoplasmic Aggregation of Truncated TDP-43
To clarify the involvement of LUBAC activity in the formation of TDP-43 aggregates, we constructed Hoip-knockout (KO) Neuro2a cells by the CRISPR/Cas9 technique, resulting in the elimination of HOIP and reduced the HOIL-1L and SHARPIN levels ( Figure S3A,B). When full length A315T 1-414 -GFP was expressed in the Hoip-KO cells, nuclear localization and non-aggregate formation were detected, as in the parental Neuro2a cells ( Figure 4A). In contrast, the number of cytosolic aggregates of A315T 247-414 -GFP were reduced in the Hoip-KO cells compared to those in the parental cells, and indeed, the intensity and size of the A315T 247-414 -GFP aggregates were significantly reduced in the Hoip-KO cells ( Figure 4B,C). Furthermore, the amount of A315T 247-414 -GFP in the RIPA-insoluble/urea-soluble aggregate fraction was reduced in the absence of Hoip, although almost equal numbers of full length A315T 1-414 -GFP were detected in both the parental and Hoip-KO Neuro2a cells ( Figure 4D). Importantly, M1-ubiquitin was detected in the insoluble fraction of the parental cells, but not the Hoip-KO cells. In the presence of MG-132, the accumulation of A315T 247-414 -GFP in the insoluble fraction was enhanced in both the parental and Hoip-KO Neuro2a cells, and the chain length of M1-ubiquitin seemed to be extended. To further examine the involvement of LUBAC activity, we constructed a LUBAC expression plasmid that could simultaneously express the HOIP, HOIL-1L, and SHARPIN subunits on a single plasmid ( Figure S4A,B). The restoration of the active LUBAC (LUBAC-WT) within the Hoip-KO cells recovered the M1-ubiquitin-positive aggregates of A315T 247-414 -GFP ( Figure 4E). In contrast, the restoration of LUBAC-CA, in which the active Cys879 of mouse HOIP was replaced by Ala, did not generate M1-ubiquitin or A315T 247-414 -GFP aggregates. For biochemical confirmation, we performed the immunoblotting of A315T-GFPs in the RIPA-soluble and RIPA-insoluble/urea-soluble fractions ( Figure 4F). Although the expression of A315T 1-414 -GFP in the RIPA-insoluble/urea-soluble fraction was not affected, regardless of the LUBAC activity in the parental and Hoip-KO Neuro2a cells, the amount of A315T 247-414 -GFP in the aggregate fraction was enhanced in the presence of LUBAC-WT, but not LUBAC-CA, in the parental and Hoip-KO Neuro2a cells. These results strongly indicate that the LUBAC-mediated linear ubiquitination activity is involved in the formation of cytoplasmic aggregates of the truncated TDP-43 mutant. Therefore, we hypothesized that the inhibition of LUBAC activity may suppress the aggregation of TDP-43.

Discussion
ALS is a fatal neurodegenerative disease. Analyses of fALS have revealed the causative genes of ALS such as TARDBP (which encodes TDP-43), SOD1, FUS, C9ORF72, ATXN2, OPTN, VCP, UBQLN2, UBQLN4, SQSTM1, TBK1, and so on, but the mechanism of ALS onset has remained elusive. TDP-43-containing cytoplasmic inclusions are reportedly detected in 97% of ALS and 45% of FTD cases [29]. Therefore, TDP-43 inclusions are a hallmark of ALS/FTD-mediated proteinopathy [3,4]. The recent structural analysis of TDP-43 282-360 showed that the region has a unique double-spiral-shaped fold, and is oligomerized with a prion-like filamentous structure [10]. Ubiquitination as well as phosphorylation is a crucial post-translational modification of TDP-43 in the brain inclusions of ALS patients [11,30]. TDP-43 contains 20 Lys residues, and multiple ubiquitination sites have been identified [31,32]. Importantly, Scotter et al. showed that the TDP-43 aggregates included both the K48-and K63-linked ubiquitin chains, which function in proteasomal degradation and autophagic degradation, respectively [33]. Furthermore, Yin et al. reported the age-dependent increase in the K63-ubiquitination of the C-terminal fragment of TDP-43, with binding to proteasome assembly proteins, PSM2 and PSD13 [34]. We determined that not only K48-and K63-linked ubiquitin chains, but also M1-linked linear ubiquitin, which regulates NF-κB activation and the cell death pathway, are involved in inclusions from OPTN-associated fALS and sALS patients, and in tau neurofibrillary tangles from Alzheimer's disease patients [20,22,23]. Although K48-ubiquitin was detectable in tiny inclusions, K63-and M1-positive inclusions were observed in K48-positive thick inclusions. The M1-ubiquitin was also colocalized with protein aggregates formed by the overexpression of pathogenic huntingtin-derived polyglutamine proteins and ataxin-3 (Machado-Joseph disease) [35]. These results suggest that the ubiquitin chains of neurodegenerative disease-associated inclusions become progressively more complex.
In this study, we showed that cytoplasmic aggregates of ectopically expressed, truncated ALS-associated TDP-43 mutants in Neuro2a cells contained M1-, K48-, and K63ubiquitins in vitro, and the MG-132-mediated inhibition of proteasome activity enhanced the ubiquitination of TDP-43 (Figures 1-3). These results indicate that TDP-43 aggregates include multiple types of ubiquitin linkages in Neuro2a cells, and may form branched and/or hybrid ubiquitin chains. Importantly, the genetic ablation of Hoip or treatment with the LUBAC inhibitor HOIPIN-8 drastically reduced the insoluble aggregates of truncated TDP-43 (Figures 4 and 5). These results suggest that the conjugation of M1-ubiquitin to TDP-43 promotes the formation of insoluble aggregates. Indeed, M1-ubiquitin reportedly has a greater propensity to form fibrillar aggregates than K48-and K63-linked ubiquitin chains in vitro [36]. Interestingly, LUBAC seemed to be localized at the peripheries of the TDP-43 aggregates, and bound to TDP-43 in an E3 activity-dependent manner ( Figure 6). We further detected the possible covalent linear ubiquitination of truncated TDP-43, and HOIPIN-8 suppressed the complex K48/K63/M1-ubiquitination of TDP-43. Since M1ubiquitination is prohibited by N-terminally-tagged ubiquitin such as His 6 -tagged ubiquitin, the conjugation of M1-ubiquitin to TDP-43 may have been overlooked in previous in vitro mass spectrometry analyses. Thus, linear ubiquitin seems to play an indispensable role in the development of various neurodegenerative diseases through the promotion of liquid-liquid phase separation (LLPS), oligomerization, and aggregate formation.
Currently, a glutamic neurotransmission inhibitor, riluzole, and an antioxidant drug, edaravone, are approved for ALS treatment [37]. Although these drugs prolong the survival of ALS patients, their efficacies are limited. Gene therapy clinical trials using antisense oligonucleotides (ASOs), RNA interference, adeno-associated virus (AAV)-mediated small hairpin RNA (shRNA), and antibody-based methods are currently being conducted [38]. Among them, the administration of ASOs targeting ataxin-2 reportedly reduced the TDP-43-positive aggregates and increased survival [39]. Thus, these gene therapies may be effective as disease-modifying treatments for ALS. In this study, we showed that aggregateprone TDP-43 truncation mutants increased the TNF-α-mediated NF-κB activity, and this was suppressed by HOIPIN-8 (Figure 7), suggesting that LUBAC activity suppression by HOIPIN-8 or its derivatives will be a potential therapeutic target to suppress ALSmediated neuroinflammation. Indeed, we have previously shown that HOIPIN-1 and -8 effectively suppressed activated B cell-like diffuse large B cell lymphoma (ABC-DLBCL) and psoriasis in mice models due to the inhibition of excessive NF-κB activity [26]. Our cellular analysis did not show a significant difference between WT 247-414 -GFP and A315T 247-414 -GFP, indicating that the production of the C-terminal fragment generally contributed to the aggregate formation, regardless of the single amino acid mutation of A315T. However, the human TDP-43 A315T mutation is a known etiology of ALS [40,41], and transgenic mice expressing the ALS-associated human A315T-mutant of TDP-43 reportedly exhibit ALS-like phenotypes with ubiquitin-positive inclusions [42,43]. Further in vivo investigations of the pharmacological and phenotypical effects of HOIPINs on ALS mice models are awaited.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.