Ubiquitin-specific peptidase-19 links TDP-43 aggregation to ER stress

Significance

The pathological hallmark of FTLD-TDP and ALS is the cytoplasmic deposition of TDP-43, a process dependent on the ubiquitin-proteasome system. Further, ER stress promotes cytoplasmic TDP-43 aggregation, which in turn, reciprocally induces ER stress. However, underlying factors critical for sustaining this pathogenic link have been a mystery. Here, we show that USP19 removes ubiquitin conjugates from TDP-43, which preferentially promotes cytoplasmic aggregation of TDP-CTFs through its catalytic activity and ER localization. USP19, in turn, sustains the coupling of TDP-CTF aggregation to ER stress, providing a feed-forward mechanism of pathogenesis. This establishes a critical role of USP19 at the nexus of TDP-43 proteostasis and ER stress, and offers a new and unique therapeutic strategy for TDP-43 proteinopathies.

Abstract

Aggregation and deposition of TAR DNA-binding protein 43 (TDP-43) is a salient pathological signature of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration-TDP (FTLD-TDP). TDP-43 proteostasis and aggregation are controlled by several posttranslational modifications, including ubiquitination. While multiple E3 ubiquitin ligases are known to facilitate TDP-43 clearance, little is known about the role of deubiquitinases (DUBs) in controlling TDP-43 proteostasis. Through an unbiased discovery screen of DUBs, here we identify and demonstrate using in vitro and in vivo models, as well as human brain tissue, that ubiquitin-specific peptidase-19 (USP19) acts as a TDP-43-directed DUB that removes K48- and K63-linked ubiquitin conjugates from TDP-43 and preferentially promotes cytoplasmic aggregation of TDP-43 C-terminal fragments (TDP-CTFs) through its catalytic activity. Specifically, the endoplasmic reticulum (ER)-anchored USP19 isoform (USP19-ER) exhibits superior activity in deubiquitinating TDP-CTFs, enhancing its phase separation and aggregation, compared to its cytosolic isoform (USP19-Cyto). Furthermore, as TDP-CTFs are generated at the ER, USP19 acts to couple the aggregation of TDP-CTFs to ER stress (ATF6, ATF4, IRE1, & CHOP). In humans, USP19 protein levels increase in FTLD-TDP brains, which extensively colocalize with cytoplasmic phospho-TDP-43 (pTDP-43) pathology. Importantly, we demonstrate in vivo that genetic reduction of usp19 mitigates pTDP-43 pathology, astrogliosis, and ER stress while reversing long-term potentiation (LTP) and motor deficits in a mouse model of TDP-43 pathogenesis (TAR4 mice). These findings establish a critical role of USP19 at the nexus of TDP-43 proteostasis and ER stress, implicating its pathogenic role in FTLD-TDP and ALS.

TAR DNA-binding protein 43 (TDP-43), a nuclear protein involved in the regulation of RNA processing, mislocalizes to the cytoplasm and aggregates in multiple neurodegenerative diseases, including frontotemporal lobar degeneration-TDP (FTLD-TDP), amyotrophic lateral sclerosis (ALS), and >50% of late-onset Alzheimer’s disease (AD) (1–3). Abnormal TDP-43 mislocalization and accumulation is associated with endoplasmic reticulum (ER) stress (4, 5), mitochondrial dysfunction (6–11), loss of synaptic integrity (12, 13), and cognitive and motor impairments (14). Notably, ectopic expression of TDP-43 induces ER stress (4, 15, 16), which activates the ER-anchored caspase-4 that cleaves TDP-43 to initiate the generation of TDP 25 kDa and 35 kDa C-terminal fragments (CTFs: TDP-25 & TDP-35) (17). While these TDP-CTFs are normally targeted for rapid proteasomal degradation (17, 18), they nevertheless accumulate and deposit in the brains of FTLD-TDP and ALS patients (1, 2). At present, the mechanism by which this occurs remains poorly understood.

TDP-43 proteostasis and aggregation are controlled by various posttranslational modifications, including phosphorylation (19–21), poly ADP-ribosylation (22, 23), oxidation (24–26), acetylation (27, 28), sumoylation (29, 30), and ubiquitination (1, 31). Recent evidence also supports a role for acetylation and/or ubiquitination in the control of TDP-43 nuclear import (32). Nonetheless, ubiquitination is a final key modification required for the turnover of TDP-43 via the ubiquitin–proteasome and autophagy–lysosome pathways. Specifically, TDP-43 ubiquitination by Parkin, Znf179, and PJA1 facilitates TDP-43 clearance (33–36), although some forms of ubiquitination, particularly under conditions of proteasome or autophagy dysfunction, may also contribute to the insolubility of target proteins (1, 37, 38). Ubiquitination is a dynamic and reversible process, as specific deubiquitinases (DUBs) can cleave ubiquitin from target proteins (39). At present, however, comparatively little is known about the role of ubiquitin removal from TDP-43 by DUBs in the brain.

The human genome encodes ~100 DUBs (40), of which ubiquitin-specific peptidases (USPs) represent the largest DUB family comprising ~50 members in humans (41). Of these, at least 20 are expressed in the central nervous system (CNS), with molecular weights ranging from 50 to 300 kDa (41, 42). Each USP is thought to have unique properties with different subsets of targets and binding partners that can process the cleavage of different types of ubiquitin chains (i.e., K6, K11, K29, K48, & K63) via its cysteine protease domain. Thus far, only 1 DUB is known to directly regulate TDP-43 ubiquitination. Specifically, overexpression of USP8 facilitates removal of ubiquitin from TDP-43 in HEK293 cells, and RNAi-mediated knockdown of USP8 in Drosophila decreases TDP-43 solubility and enhances its toxicity (43), indicating that this form of TDP-43 deubiquitination mediated by USP8 increases TDP-43 solubility and decreases its toxicity. To investigate the broader role of USPs in regulating TDP-43, we executed an siRNA screen against 20 USPs expressed in the CNS and identified USP19 as the strongest candidate as a TDP-43-directed USP. Because a major splice variant of USP19 localizes to the ER (44, 45), we hypothesized that it might contribute to both TDP-43 proteostasis and associated ER stress. Here, we present both in vitro and in vivo evidence, including in human brain tissue, that USP19-mediated deubiquitination of TDP-43 enhances its pathological deposition, contributing to TDP-43-induced ER stress, loss of synaptic integrity, long-term potentiation (LTP) deficits, and motor impairment.

Results

Identification of USP19 as a TDP-43-Directed Deubiquitinase.

We carried out an siRNA screen of USPs known to be expressed in the mammalian CNS in HeLa-GFP-TDP-43 cells stably expressing tetracycline-inducible human wild type GFP-TDP-43 (46) using Smartpool siRNAs targeting 20 different USPs. To quantitatively estimate ubiquitin conjugation on TDP-43, we used the Duolink^® proximity ligation assay (PLA, Fig. 1A, schematic) for TDP-43 and ubiquitin association using antibodies directed against each protein, which detects TDP-43 in close proximity with ubiquitin (~10 Å) in situ (red puncta) (47) (Fig. 1B). Among the targeted USPs, USP19 siRNA significantly increased ubiquitin–TDP-43 association as measured by the numbers of PLA puncta per cell (Fig. 1 B and C), suggesting that USP19 reduction enhances TDP-43 ubiquitination. Knockdown of USP7, USP54, and USP8, the latter a DUB previously shown to deubiquitinate TDP-43 (43), also increased ubiquitin–TDP-43 PLA puncta number per cell without reaching statistical significance (Fig. 1C). While knockdown of USP13, USP22, or USP40 modestly decreased ubiquitin–TDP-43 PLA puncta numbers per cell compared to the control (Fig. 1C), they did not reach statistical significance.

Fig. 1.

Identification of USP19 as a TDP-43-directed deubiquitinase. (A) Schematic of Duolink® PLA principle for detection of ubiquitin (Ub)-TDP-43 complexes. (B) Representative images of Ub-TDP-43 PLA puncta (red), TDP-43 (green), and DAPI (blue) in tetracycline (tet)-induced HeLa-GFP-TDP-43 cells transfected with Smartpool siRNAs targeting 20 different USPs. (C) Quantification of Ub-TDP-43 PLA puncta numbers per cell (one-way ANOVA, F(20, 42) = 2.711, P = 0.0032; post hoc Dunnett compared to control siRNA (Cont), *P = 0.0270; n = 3 independent experiments per USP with 4 to 9 images averaged per experiment). (D) Schematic of USP19-flag constructs: ER-localized USP19 (USP19-ER), cytosolic USP19 (USP19-Cyto), and USP19-ER lacking catalytic DUB activity (USP19-CS-ER). (E) Representative blots showing co-IP of Myc-TDP-43 with USP19-flag variants (flag pulldown) in HEK293T cells cotransfected ± USP19-flag variants or vector control ± Myc-TDP-43. (F) Representative images of USP19–TDP-43 PLA puncta (red: myc + flag PLA) and DAPI (blue) from HeLa cells transfected with Myc-TDP-43 and USP19-flag variants or vector control. One probe - lacks one secondary antibody probe. (G) Representative blots of polyubiquitin conjugates (Ub-K48 & Ub-K63) in TDP-43 pulldown (myc) from HEK293T cells transfected with ubiquitin-HA (Ub-HA) and myc-TDP-43 with USP19-flag variants or vector control.

USP19 protein is expressed in the CNS in both neurons and glia (48, 49), whereas USP54 protein expression in the brain has not yet been examined. Moreover, USP19 presented as the strongest candidate for a TDP-43-directed USP. Thus, we next focused on USP19 to validate our screening results. USP19 contains two p23/CS domains within the N-terminal region that are found in cochaperones to Hsp90, p23, and Sgt1 (45). As USP19 binds Hsp90 and Hsp70/Hsc70, the p23/CS domains may serve in a cochaperone capacity (45). USP19 also contains a large C-terminal catalytic domain with distantly spaced Cys, Asp, and His residues required for catalytic DUB activity (Fig. 1D). Embedded within the large catalytic domain are Ubl and MYND Zn finger (Znf) containing domains, involved in protein–protein interactions (50). USP19 is expressed as two major isoforms arising from alternative splicing of the final exon. Use of exon 27 adds a transmembrane (TM) segment to the C-terminus that targets USP19 for insertion into the ER membrane with the catalytic domain facing the cytosol (USP19-ER) (Fig. 1D and SI Appendix, Fig. S1, ER localization), whereas use of exon 28 excludes the TM segment, which renders USP19 cytosolic (USP19-Cyto) (Fig. 1D) (44, 45).

To determine if USP19 variants differentially form complexes with TDP-43, we transfected HEK293T cells with myc-TDP-43 and vector control, USP19-ER-flag, USP19-Cyto-flag, or the catalytically dead USP19-CS-ER-flag (Fig. 1D). Myc-TDP-43 was detected in Flag immune complexes of all 3 USP19-Flag variants in co-IP experiments (Fig. 1E). Likewise, TDP-43 formed complexes with all three USP19 variants as detected by in situ Duolink^® PLA (Fig. 1F). To determine if USP19 deubiquitinates TDP-43, we transfected HEK293T cells with myc-TDP-43 + ubiquitin-HA together with vector control, USP19-ER-flag, USP19-Cyto-flag, or USP19-CS-ER-flag, followed by TDP-43-myc pulldown and detection of K48- or K63-linked ubiquitin (Ub-K48 or Ub-K63). Indeed, both the ER and cytosolic forms of USP19 dramatically reduced high molecular weight Ub-K48 and Ub-K63 on TDP-43 (Fig. 1G). By contrast, the catalytically dead USP19-CS-ER failed to reduce Ub-K48 or Ub-K63 conjugates on TDP-43 (Fig. 1G), indicating that USP19 DUB activity is essential in removing Ub-K48 and Ub-K63 conjugates from TDP-43. Together, these results validate the identification of USP19 as a TDP-43-directed DUB.

USP19 Promotes the Accumulation of TDP-CTFs.

We next used HeLa-GFP-TDP-43 stable cells (46) to investigate the role of endogenous USP19 in TDP-43 proteostasis. Knockdown of USP19 did not significantly alter the 80-kDa exogenous GFP-TDP-43 (TDP-80), 43-kDa endogenous TDP-43, or 35-kDa TDP-CTF35 (TDP-35) fragment in the RIPA-soluble fraction (Fig. 2 A–D). However, USP19 knockdown reduced exogenous TDP-80 by ~30% and TDP-35 by ~50% in the RIPA-insoluble fraction (Fig. 2 A, B, and D). As the TDP-35 fragment appeared to be preferentially reduced by USP19 knockdown, we transiently transfected TDP-35-myc in the high-efficiency transfection cell line HEK293T with or without USP19 siRNA. Here, TDP-35 and the proteolytically cleaved 25-kDa TDP-25 fragment were observed in control siRNA transfected cells (Fig. 2E). By contrast, USP19 knockdown reduced TDP-35 and TDP-25 by ~50% (Fig. 2 F and G), the latter a CTF that accumulates in human TDP proteinopathies (1, 2). We next utilized primary neurons derived from TAR4^+/+ mice expressing wild-type human TDP-43, coupled with adenovirus-mediated expression of USP19-ER (Fig. 2H). Ectopic USP19-ER expression did not alter TDP-43 either in the RIPA-soluble or RIPA-insoluble fraction (Fig. 2 H–J). In contrast, USP19-ER overexpression dramatically increased RIPA-insoluble TDP-35 by ~4.3-fold (Fig. 2 H, K, and L), indicating that USP19-ER promotes the conversion of soluble TDP-35 to an insoluble state.

Fig. 2.

USP19 drives TDP-43 aggregation by increasing insoluble TDP-43 CTFs. (A) Representative blots of RIPA-soluble and RIPA-insoluble proteins from tet-induced HeLa-GFP-TDP-43 cell line transfected with control or USP19 siRNA. (B–D) Quantification of soluble (sol) and insoluble (insol) TDP-80 (GFP-TDP-43), TDP-43, and TDP-35 from RIPA-soluble and RIPA-insoluble fractions (one-sample t test, *P < 0.05, **P < 0.01, n = 4 independent experiments with 2 to 3 replicates averaged per experiment). (E) Representative RIPA-soluble blots of HEK 293T cells transfected with TDP-35-myc and control siRNA or USP19 siRNA. (F and G) Quantification of TDP-35 and TDP-25 (one-sample t test, *P < 0.05; n = 4 independent experiments with 2 to 3 replicates averaged per experiment). (H) Representative blots of RIPA-soluble and RIPA-insoluble proteins from TAR4^+/+ primary neurons transduced with control or USP19 adenovirus. (I–L) Quantification of TDP-43 and TDP-35 from RIPA-soluble and RIPA-insoluble fractions (one-sample t test, *P < 0.05, ns: not significant; n = 2 independent experiments with two replicates averaged per experiment).

The USP19-ER Isoform Preferentially Deubiquitinates and Increases Insoluble TDP-CTF through Phase Separation.

To determine whether USP19 variants differentially alter TDP-43, we transfected HeLa-GFP-TDP-43 cells with vector control, USP19-ER, USP19-CS-ER, or USP19-cyto. None of the USP19 variants significantly altered RIPA-soluble exogenous GFP-TDP-80, endogenous TDP-43, or TDP-35 fragment (Fig. 3 A–D). In the RIPA-insoluble fraction, USP19-ER increased TDP-35 by a significant ~2.5-fold (Fig. 3 A and G) but did not significantly alter exogenous GFP-TDP-80 or endogenous TDP-43 (Fig. 3 A, E, and F). By contrast, neither USP19-CS-ER nor USP19-Cyto significantly altered any of the TDP bands in the RIPA-insoluble fraction (Fig. 3 A and E–G), indicating that USP19 DUB activity and ER localization are critical for increasing insoluble TDP-35. As TDP-43 induces ER stress (4, 15, 16) and activates ER-anchored caspase-4 that cleaves TDP-43 to initiate the generation of TDP-25 and TDP-35 fragments (17), we wondered whether ER localization of USP19 might preferentially promote the deubiquitination of TDP-35. Indeed, USP19-ER demonstrated markedly superior efficacy in deubiquitinating TDP-35 than USP19-Cyto (Fig. 3H), indicating that USP19 tethering to the ER enhances TDP-35 deubiquitination.

Fig. 3.

USP19-ER isoform preferentially deubiquitinates and increases insoluble TDP-CTF through phase separation. (A) Representative blots of RIPA-soluble and RIPA-insoluble proteins from tet-induced GFP-TDP-43 cells transfected with vector control or USP19 variants. (B–G) Quantification of TDP-80 (GFP-TDP-43), TDP-43, and TDP-35 from RIPA-soluble (sol) and RIPA-insoluble (insol) fractions (Insol TDP-35: one-way ANOVA, F(3, 8) = 6.108, P = 0.0183; post hoc Dunnett, **P < 0.01; ns: not significant; n = 3 independent experiments with 1 to 2 replicates averaged per experiment). (H) Representative blots of polyubiquitin conjugates (Ub-K48 & Ub-K63) in myc-TDP-35 pulldown (myc) from HEK293T cells transfected with ubiquitin-HA (Ub-HA) and myc-TDP-35 with USP19-flag variants or vector control. (I) Schematic model blue light-induced phase separation of Cry2olig-TDP-LCD-mCherry. (J) Representative images of Cry2olig-TDP-LCD-mCherry T(red) before and after blue light stimulation (488 nm, 1 s) in HeLa cells transfected with Cry2olig-TDP-LCD and USP19 variants (green) or vector control. (K) Quantification of Cry2olig-TDP-LCD puncta numbers normalized to cell area (two-way ANOVA, F(3, 39) = 21.33, P < 0.0001; post hoc Dunnett compared to control, ****P < 0.0001, ns: not significant; n = 2 independent experiments with 2 to 3 replicates averaged per experiment).

The C-terminus of TDP-43 contains a glycine-rich low complexity domain (LCD), a region harboring the majority of ALS/FTD-causing missense mutations (51). LCDs of multiple RNA-binding proteins, including TDP-43 mediate protein and RNA interactions through liquid–liquid phase separation (LLPS) (51). Aberrant or excessive TDP-43 phase separation is thought to drive the formation and maturation of pathological inclusions (51, 52). The optogenetic Cry2olig-TDP-LCD-mCherry reporter was recently used to model TDP-43 phase separation and proteinopathy under spatiotemporal control of light stimulation (53). While a brief pulse of blue light (465 nm) induces the oligomerization of Cry2olig-mCherry and is quickly reversed, the same light pulse induces the oligomerization and phase separation of Cry2olig-TDP-LCD-mCherry, which is less reversible and progressively becomes more irreversible with repeated stimulations (Fig. 3I) (53). We cotransfected Cry2olig-TDP-LCD-mCherry with USP19-ER, USP19-CS-ER, USP19-Cyto, or vector control in NIH3T3 cells, a flat cell line amenable to live-cell fluorescence imaging. By live cell imaging, we pulsed cells for 1 s with 465 nm blue light and imaged cells over 300 s. Indeed, USP19-ER significantly increased TDP-LCD phase separation compared to vector control as detected by increased TDP-LCD puncta numbers, which peaked at 60 s and persisted over 300 s (Fig. 3 J and K). By contrast, USP19-CS-ER and USP19-Cyto failed to significantly increase TDP-LCD puncta numbers (Fig. 3 J and K), indicating that USP19 DUB activity accelerates TDP-LCD phase separation at ER sites. While USP19-ER also demonstrated the most robust and significant increase in TDP-LCD puncta area, which peaked at 60 s, USP19-Cyto but not USP19-CS-ER significantly increased TDP-LCD puncta area (SI Appendix, Fig. S2), indicating that USP19-Cyto has the capacity to promote cytosolic TDP-LCD phase separation, possibly in a manner distinct from USP19-ER.

USP19 Promotes Cytoplasmic Mislocalization of TDP-43 and Enhances ER Stress.

To determine whether endogenous USP19 facilitates cytoplasmic mislocalization of TDP-43, we transfected HeLa cells with TDP-43-tomato and control siRNA or USP19 siRNA and treated cells with the proteasome inhibitor MG132 for 4 h, which promotes cytoplasmic mislocalization of TDP-43 (4). While control siRNA transfected cells showed prominent cytoplasmic TDP-43-tomato mislocalization with MG132 treatment, USP19 siRNA strongly reduced the ratio of cytoplasmic to nuclear TDP-43-tomato by ~65% (Fig. 4 A and B). By contrast, overexpression of USP19-ER but not the catalytically dead USP19-CS-ER significantly increased the ratio of cytoplasmic to nuclear TDP-43-tomato by ~twofold compared to GFP control in the absence of MG132 treatment (Fig. 4 C and D), indicating that the catalytic activity of USP19 is required to drive cytoplasmic TDP-43 accumulation. Similar results for USP19-mediated mislocalization of endogenous TDP-43 were observed in HeLa cells lacking exogenous TDP-43 expression (SI Appendix, Fig. S3 A–D).

Fig. 4.

USP19 promotes cytoplasmic mislocalization of TDP-43 and enhances ER stress. (A) Representative images of TDP-43-tomato (red), USP19 (green), and DAPI (blue) in HeLa cells transfected with TDP-43 and control siRNA or USP19 siRNA ± proteasome inhibition with MG132 treatment (10 μM, 4 h). (B) Quantification of TDP-43-tomato in the cytoplasm vs. nucleus with MG132 treatment (one-sample t test, *P < 0.05; n = 3 independent experiments with 10 to 32 cells/condition averaged per experiment). (C) Representative images of TDP-43-tomato (red), USP19 (green), and DAPI (blue) in HeLa cells transfected with TDP-43 and USP19 variants or GFP control. (D) Quantification of TDP-43-tomato in the cytoplasm vs. nucleus (one-way ANOVA, F(2, 6) = 22.73, P = 0.0016; post hoc Dunnett, **P < 0.005, n = 3 independent experiments with 13 to 33 cells/condition averaged per experiment). (E) Representative images of endogenous USP19–TDP-43 PLA puncta (green) and DAPI (blue) in HeLa cells transfected with control or USP19 siRNA. Negative controls with only 1 primary antibody or 1 secondary antibody probe (Bottom). (F) Representative images of endogenous USP19–TDP-43 PLA puncta (red) and DAPI (blue) in HeLa cells treated ± ER stressor tunicamycin (2 μM) for 24 h. Negative controls with only 1 primary antibody or 1 secondary antibody probe (Bottom). (G) Quantification of USP19–TDP-43 PLA intensity ± tunicamycin treatment (2 μM, 24 h) (one-sample t test, ***P < 0.001, n = 3 independent experiments with 29 to 36 cells/condition averaged per experiment). (H) Schematic of unfolded protein response (UPR) triggered by ER stressors, such as misfolded proteins, illustrating the activation of IRE1α-XBP1, PERK-ATF4, and ATF6 pathways, leading to CHOP induction. (I) Representative blots of ER stress-induced UPR markers from tet-induced HeLa-GFP-TDP-43 cells transfected with control or USP19 siRNA and treated ± tunicamycin (2 μM; 0, 8, 24 h). (J–M) Quantification of CHOP, cleaved ATF6 (cl-ATF6), ATF4, and IRE1α (two-way ANOVA; CHOP: F(1, 18) = 24.49, P = 0.0001; cl-ATF6: F(1, 18) = 41.93, P < 0.0001; ATF-4: F(1, 18) = 4.343, P = 0.0517; IRE1α: F(1, 18) = 2.80, P = 0.1571; post hoc Sidak, *P < 0.05, **P < 0.005, ***P < 0.001, ns: not significant; n = 4 independent experiments with 2 to 3 replicates averaged per experiment).

We next tested whether ER stress induces greater association of endogenous TDP-43 with endogenous USP19. To detect USP19–TDP-43 protein complexes, we transfected HeLa cells with control siRNA or USP19 siRNA, and subjected cells to PLA to detect endogenous USP19–TDP-43 complexes in situ. Indeed, we observed endogenous USP19–TDP-43 PLA puncta, which diminished in cells transfected with USP19 siRNA (Fig. 4 E, Top). Negative controls (1 probe or 1 primary antibody) yielded no PLA signal (Fig. 4 E, Bottom). Treatment of HeLa cells with the ER stressor tunicamycin for 24 h, which activates the ER unfolded protein response (UPR^ER), significantly increased endogenous USP19–TDP-43 complexes by ~twofold (Fig. 4 F and G). Notably, most of the complexes were localized to the cytoplasm (Fig. 4F), suggesting that USP19–TDP-43 complexes form upon TDP-43 mislocalization to the cytoplasm under proteotoxic stress. As expected, a similar increase in USP19–TDP-43 association was seen after tunicamycin treatment in HeLa-GFP-TDP-43 cells (SI Appendix, Fig. S4A).

Excessive accumulation of misfolded proteins in the ER initiates the UPR^ER by activating IRE1, PERK-ATF4, and ATF6, resulting in transcription of downstream genes, including CHOP (Fig. 4H) (54). Previous studies have shown that ectopic expression of TDP-43 induces ER stress (4, 15, 16), which activates the ER-anchored caspase-4 that initiates the cleavage of TDP-43 to its C-terminal fragments (TDP-25 & TDP-35) (17). Given that ER-anchored USP19 regulates ER-associated degradation (ERAD) (45) and increases insoluble TDP-CTFs, we investigated whether USP19 regulates ER stress and UPR^ER in the absence and presence of ectopic TDP-43 expression. We transfected parental HeLa cells or HeLa-GFP-TDP-43 cells with control or USP19 siRNA and treated cells with the ER stress inducer tunicamycin for the indicated times. Cells were then subjected to Western blotting for detection of UPR^ER markers, IRE1α, ATF4, cleaved ATF6 (cl-ATF6), and CHOP. In parental HeLa cells, USP19 knockdown moderately reduced tunicamycin-induced UPR^ER markers (ATF4 and IRE1α) but failed to blunt cleaved ATF6 or CHOP induction (SI Appendix, Fig. S5 A–E), indicating a modest level of ER stress suppression by USP19 knockdown through ATF4 and IRE1α pathways. In HeLa-GFP-TDP-43 cells with ectopic TDP-43 expression, however, USP19 siRNA robustly blunted the induction cleaved ATF6 and CHOP at both 8 h and 24 h of tunicamycin treatment (Fig. 4 I–K). Although USP19 knockdown also moderated ATF4 and IRE1α induction in HeLa-GFP-TDP-43 cells, this only reached significance at 24 h (Fig. 4 I, L, and M). We also treated HeLa-GFP-TDP-43 cells with another ER stressor, thapsigargin. Staining for CHOP demonstrated that USP19 siRNA likewise reduces thapsigargin-induced CHOP levels by ~50% (SI Appendix, Fig. S4 B and C). Thus, the ATF6-CHOP branch of UPR^ER appears to be primed by TDP-43 overload and USP19 expression, rendering this pathway more sensitive to TDP-43 and USP19 levels.

Elevated Levels of USP19 in FTLD-TDP Brains Colocalize with pTDP-43 Pathology.

To determine whether USP19 levels are altered in human TDP-43 proteinopathy, we assessed USP19 levels by Western blotting from RIPA-soluble fractions of human FTLD-TDP patients and age-matched nondementia controls (SI Appendix, Fig. S6, case information). USP19 levels normalized to tubulin were significantly elevated by >threefold in FTLD-TDP brains compared to controls (Fig. 5 A and B). We next double stained for cytoplasmic pTDP-43 pathology (pS403/pS404-TDP-43, green) and USP19 (red) in the frontal cortex to assess their potential colocalization. Indeed, pTDP-43-positive inclusions were widespread in FTLD-TDP brains (Fig. 5C1, Upper, green), and USP19 extensively colocalized with pTDP-43 inclusions (Fig. 5C1, Lower, merged) with Pearson’s colocalization coefficient of r = 0.623. Staining with only secondary antibodies failed to detect any signal (Fig. 5C2, 3 Right). In nondementia controls, we observed little to no pTDP-43 pathology (Fig. 5C2, 3 Upper Left, green), whereas cytoplasmic and neuropil USP19 staining was seen (Fig. 5C2, 3 Middle Left, red). These results indicate a significant role of USP19 in human TDP-43 proteinopathy and place USP19 at the scene of pathogenesis.

Fig. 5.

Elevated levels of USP19 in FTLD-TDP brains colocalize with pTDP-43 pathology. (A) Representative blots of RIPA-soluble USP19 and tubulin from the frontal cortex of FTLD-TDP and nondementia control cases. (B) Quantification of USP19 levels (unpaired Student’s t test, ***P = 0.0002, n = 5 FTLD-TDP and n = 11 nondementia controls). (C1 and C2) Representative double-stained immunohistochemical (IHC) images of phospho-TDP-43 (pS403/pS404, green) and USP19 (red) in 4 FTLD-TDP cases and three controls, showing extensive colocalization of USP19 with pTDP-43 pathology (merged) in FTLD-TDP. Secondary antibody only negative controls (C2, Right).

In Vivo Reduction of usp19 Mitigates TDP-43 Pathology, Lowers ER Stress, and Reverses Hippocampal LTP and Motor Deficits.

Hemizygous TAR4^+/− mice express wild type human TDP-43 driven by the mThy-1 promoter and exhibit pTDP-43 inclusions associated with hippocampal LTP impairment (55) and deficits in rotarod motor performance (34, 56). Usp19^+/− mice are grossly normal, adult viable, fertile, and exhibit no salient abnormalities, although male but not female usp19^−/− mice exhibit subfertility and are ~10% smaller (57). To determine whether reducing endogenous USP19 impacts TDP-43 pathology in vivo, we crossed TAR4^+/− mice with usp19^+/− mice, the latter which exhibited 50% lower USP19 protein levels compared to WT mice (SI Appendix, Fig. S7 A and B). Cytoplasmic pTDP-43 (pS409/pS410-TDP-43) inclusions were widespread in TAR4^+/− at 10 mo of age, which were significantly reduced by ~55% in TAR4^+/−;usp19^+/− mice (Fig. 6 A and B). To determine whether USP19 reduction differentially alters the levels of cytoplasmic vs. nuclear TDP-43 and TDP-35, we biochemically isolated nuclear and cytoplasmic fractions. Compared to TAR4^+/− brains, TAR4^+/−;usp19^+/− brains exhibited significant ~55 to 60% reductions in cytoplasmic TDP-43 and TDP-35 without altering nuclear TDP-43 (Fig. 6 C–F), indicating that usp19 reduction selectively impacts mislocalized cytoplasmic TDP-43. Nuclear TDP-35 was undetectable. TAR4^+/− mice exhibited a modest but significant increase in GFAP staining compared to WT mice at 10 mo of age, which was fully and significantly reversed in TAR4^+/−;usp19^+/− mice (Fig. 6 G and H). Iba1 intensity, however, was unchanged among WT, TAR4^+/−, and TAR4^+/−;usp19^+/− mice at 10 mo of age (SI Appendix, Fig. S7 C and D). Staining for the ER stress marker CHOP showed ~ threefold elevated levels of CHOP in TAR4^+/− compared to WT brains, which was nearly restored to WT levels in TAR4^+/−;usp19^+/− brains (Fig. 6 I and J), confirming that USP19 reduction mitigates TDP-43-induced ER stress in vivo.

Fig. 6.

Usp19 reduction mitigates TDP-43 pathology, lowers ER stress, and reverses LTP and motor deficits. (A) Representative IHC images of cytoplasmic pTDP-43 pathology (pS409/pS410, green) and DAPI (blue) in the cortex of 10-mo-old TAR4^+/− and TAR4^+/−;usp19^+/− mice. (B) Quantification of pTDP-43 pathology (pS409/pS410) in the cortex (unpaired Student’s t test, **P = 0.003, n = 4 mice/genotype with 4 to 10 images averaged per mouse). (C) Immunoblots of human TDP-43 and TDP-35 in cytoplasmic (Cyto) and nuclear (Nucl) fractions of TAR4^+/− and TAR4^+/−;usp19^+/− mice cortex. (D–F) Quantification of cytoplasmic and nuclear TDP-43 and TDP-35 (unpaired Student’s t test, *P < 0.05, **P < 0.01, ns: not significant; n = 3 to 4 mice/genotype). (G) Representative IHC images of GFAP (green) and DAPI (blue) in the hippocampus of 10-mo-old WT, TAR4^+/− and TAR4^+/−;usp19^+/− mice. (H) Quantification of GFAP intensity in the hippocampus (one-way ANOVA, F(2, 12) = 5.565, P = 0.0195; post hoc Dunnett, *P < 0.05; n = 4 to 6 mice/genotype with 3 to 6 images averaged per mouse). (I) Representative IHC images of CHOP (red) and DAPI (blue) in the cortex of 10-mo-old WT, TAR4^+/− and TAR4^+/−;usp19^+/− mice. (J) Quantification of CHOP intensity in the cortex (one-way ANOVA, F(2, 9) = 6.678, P = 0.0167; post hoc Dunnett, *P < 0.05, n = 4 mice/genotype with 8 to 34 images averaged per mouse). (K) Quantification of LTP induced by theta burst stimulation from 9-mo-old mouse brain slices (two-way ANOVA, F(2, 9198) = 870.6, P < 0.0001; post hoc Dunnett, ****P < 0.0001; n = 29 to 51 slices from 5 to 7 mice/genotype). (L) Quantification of latency to fall (sec) in rotarod test from 9 to 10-mo-old mice (one-way ANOVA, F(2, 51) = 18.13, P < 0.0001; post hoc Dunnett, **P < 0.005, ****P < 0.0001, n = 14 to 20 mice/genotype).

To assess potential changes in short- and long-term synaptic plasticity, we conducted electrophysiological recordings of acute brain slices from 10-mo-old mice, stimulating from the Schaffer collaterals emanating from the CA3 and recording from the CA1 of the hippocampus (SI Appendix, Fig. S7E, schematic). In recordings of paired-pulse facilitation (PPF), a form of presynaptic efficacy, we did not detect notable differences among WT, TAR4^+/−, and TAR4^+/−;usp19^+/− slices (SI Appendix, Fig. S7F). We next applied theta-burst stimulation to induce long-term potentiation (LTP). As previously documented (55), TAR4^+/− slices exhibited significantly blunted LTP over the 1 h recording period compared to WT slices (Fig. 6K). By contrast, TAR4^+/−;usp19^+/− slices exhibited LTP induction and maintenance essentially identical to those from WT slices (Fig. 6K), indicating that usp19 reduction recovers long-term synaptic plasticity in TAR4^+/− mice. PPF and LTP were indistinguishable between WT and usp19^+/− slices (SI Appendix, Fig. S7 G and H).

We next performed rotarod motor testing to determine whether usp19 reduction alleviates motor deficits seen in TAR4^+/− mice (34, 56). At 9 to 10 mo of age, TAR4^+/− mice exhibited ~65% shorter latency to fall from the rotarod compared to WT mice (Fig. 6L), indicative of motor impairment. By contrast, TAR4^+/−;usp19^+/− mice significantly restored rotarod performance to near WT levels (Fig. 6L). Interestingly, this restoration of performance by TAR4^+/−;usp19^+/− mice was more robust among male mice than among female mice (SI Appendix, Fig. S7 I and J). Different genotypes had no effect on the weight of mice (SI Appendix, Fig. S7 K and L), indicating that rotarod performance was not influenced by weight. WT and usp19^+/− mice did not differ in their rotarod performance (SI Appendix, Fig. S7M), indicating that usp19 reduction in TAR4^+/− mice restores deficits in motor performance induced by TDP-43.

Discussion

In this study, we executed a siRNA screen and identified USP19 as a TDP-43-directed DUB, which makes a significant advancement in understanding a mechanism driving TDP-43 proteinopathy. Specifically, we found that USP19 binds TDP-43, removes both K48- and K63-linked ubiquitin chains, and promotes the aggregation of cytoplasmically mislocalized TDP-43 and its C-terminal fragments (TDP-35/TDP-25) through phase separation. Our observation that ER-localized USP19 preferentially deubiquitinates and promotes the aggregation of TDP-CTFs over cytosolically localized USP19 introduces spatial regulation to TDP-43 proteostasis, which aligns with USP19’s role in enhancing ER stress. Moreover, we found that USP19 increases in brains of FTLD-TDP patients and, colocalizes with TDP-43 pathology, establishing its pathological significance in the disease setting. Accordingly, genetic reduction of usp19 mitigates TDP-43 pathology, astrogliosis, ER stress, LTP deficits, and motor impairment seen in an animal model of TDP-43 pathology. These findings, therefore, establish USP19 as a critical regulator at the nexus of TDP-43 proteostasis and ER stress, implicating its pathogenic role in TDP-43 proteinopathies.

It is notable that a prior study showed that USP8 suppresses TDP-43 insolubility and toxicity in Drosophila models, whereas we observed an opposing effect by USP19, which promotes TDP-43 insolubility and aggregation. This divergence may arise from several factors. While full-length TDP-43 exhibits a long half-life (>24 h) and is generally not targeted by the proteasome, TDP-CTFs are rapidly cleared by the proteasome (18) once generated at ER sites by caspase-4 in coordination with caspases-3/7 (17). Thus, the subcellular context of USP19’s ER localization may play a key role in its divergence from USP8. As USP19 preferentially deubiquitinates and stabilizes TDP-CTFs, it is plausible that removal of K48- and/or K63-linked ubiquitin chains from the CTFs at ER sites by USP19 impedes their degradation by the proteasomal and/or autophagic machinery, resulting in their accumulation and aggregation. Indeed, cytoplasmic TDP-43 aggregates form initially at ER sites and spread to locations beyond the ER (4, 17). As TDP-CTFs lack nuclear localization signals and contain the aggregation-prone LCD region, USP19’s activity in removing ubiquitin degradation signals creates a permissive environment for phase separation of TDP-CTFs. The resultant CTF aggregates may sequester full-length TDP-43 (17), perpetuating a feedforward cycle of cytoplasmic mislocalization and insolubility. Indeed, TDP-CTFs are more prone to aggregation and seeding (58, 59). By contrast, USP8 lacks ER anchoring and thus may preferentially process nuclear or cytosolic pools of TDP-43, rendering differential effects. These distinctions highlight the contextual specificity of DUB activity, where subcellular localization and disease associated proteolytic fragments determine pathological outcomes. Additional factors that may help to explain the divergence between USP19 and USP8 potentially include their processing of ubiquitin chains on different lysine residues and/or distinct ubiquitin chain types that induce different conformations of TDP-43. While the types of TDP-43 ubiquitin linkages cleaved by USP8 are unknown, it is notable that USP8 greatly favors the cleavage of K63-linked ubiquitin chains on other substrates (60, 61). By contrast, USP19 cleaves both K48- and K63-linked ubiquitin from TDP-43.

Previous studies have highlighted a role for TDP-43 in ER stress. Specifically, pharmacological induction of ER stress promotes cytoplasmic mislocalization of TDP-43, while TDP-43 mislocalization and aggregation reciprocally induce ER stress (4, 5, 17) (SI Appendix, Fig. S8, schematic). Indeed, TAR4^+/− mice exhibited highly elevated CHOP levels. ER stress activates UPR^ER, leading to induction of ER stress-responsive genes and inhibition of protein translation (5, 54). While this homeostatic response is initially protective and helps to resolve short-term cellular insult, prolonged and chronic ER stress leads to detrimental effects ultimately resulting in apoptotic cell death (5, 54). Our observation that USP19 knockdown blunts ER stress-induced UPR activation indicates that USP19 directly couples TDP-43 aggregation to UPR activation. This effect is plausibly mediated by USP19-mediated accumulation of insoluble TDP-CTFs, which induces ER stress and triggers UPR sensors, including ER-localized caspase-4 (SI Appendix, Fig. S8). Caspase-4 together with caspases-3/7 then further cleave TDP-43 into CTFs (17), which are again stabilized by ER-anchored USP19, thereby creating a vicious feedforward loop that sustains ER stress (SI Appendix, Fig. S8). Hence, USP19 depletion appears to decouple this pathological feedforward loop.

USP19 also regulates ERAD, particularly during ER stress. Specifically, the USP19 transcript is induced during ER stress (45), and ER-anchored USP19 rescues the ERAD substrates CFTR, TCRα, and LRP6 from proteasomal degradation (45, 62, 63). USP19 is also able to deubiquitinate and stabilize the ERAD E3 ligase HRD1 (62), thereby fine-tuning the extent of rescue from ERAD. This process pertains to TDP-43, as cleavage of TDP-43 into CTFs at the ER represents an intermediate step in a pathway to remove cytoplasmically mislocalized TDP-43 by the proteasome (17, 18). However, the ability of ER-anchored USP19 to rescue substrates from ERAD exacerbates TDP-43 proteinopathy by also rescuing its CTFs from proteasomal degradation and exacerbating ER stress. Thus, excessive USP19 activity appears to operate at the nexus of proteostatic collapse and ER dysfunction in TDP-43 proteinopathies. In this light, it is intriguing that the ATF6-CHOP branch of UPR appears to be primed by ectopic TDP-43, which becomes highly sensitive to USP19 knockdown. As TDP-43 promotes both ATF6 and CHOP induction (4, 64), we speculate that cytoplasmic TDP-43 overload together with USP19 cooperatively primes the ATF6 branch of UPR through TDP-CTF accumulation at the ER membrane, akin to selective ATF6 activation induced by stress at the ER bilayer (65, 66). This also suggests that USP19 enhances cytoplasmic gain of TDP toxicity rather than directly impacting the loss of TDP-43 nuclear function (67).

Beyond USP19 as a TDP-43 directed DUB identified in this study, a previous study showed that USP19 increases the protein level and aggregation of polyQ-expanded proteins such as Ataxin-3 and Huntingtin via its interaction with Hsp90, suggesting that USP19 may participate in triage decisions for proteostasis of Hsp90 substrates (68). Another proposed role for USP19 is in an unconventional secretion mechanism dubbed “misfolding-associated protein secretion” (MAPS), in which misfolded proteins (i.e., α-synuclein) are deubiquitinated by ER-anchored USP19 and encapsulated into ER-associated late endosomes for secretion of out of cells (69, 70). Indeed, loss of USP19 mitigates the accumulation of phospho-α-synuclein aggregates in a preformed fibril (PFF) mouse model of Parkinson’s disease (αSyn^A53T), possibly through MAPS (48). Further, a recent study showed that USP19-ER strongly promotes the secretion of the ALS-linked TDP-43-K263E mutant through MAPS, whereas USP19-ER had little effect on the secretion of wild type TDP-43 (71). Whether such secretory action contributes to TDP pathology in TAR4 mice expressing wild type TDP-43 remains speculative. In the current study, genetic reduction of USP19 mitigated cytoplasmic TDP-43 pathology and lowered ER stress together with reversal of LTP and motor deficits in TAR4 mice. These results therefore position USP19 as a promising therapeutic target at the nexus of TDP-43 proteinopathies and ER stress, with strong implications for other neurodegenerative conditions, and joins a family of DUBs playing pivotal roles in distinct neurodegenerative pathologies (72, 73).

Materials and Methods

DNA Constructs, siRNA, and Adenovirus.

The following plasmids were obtained from addgene: USP19-ER (#78597), USP19-Cyto (#78579), USP19-CS-ER (#78584), HA-Ubiquitin (#18712), USP19-ER-mCitrine (#78593), USP19-Cyto-mCitrine (#78595), USP19-CS-mCitrine (#78594), TDP-43-tomato-HA (#28205), and Cry2olig-mCherry (#60032). pcDNA (Invitrogen, Cat# V79020) was purchased from Invitrogen. Myc-TDP-43, myc-TDP-35, Cry2olig-TDP-LCD-mCherry constructs were cloned in house. ON-TARGET smartpool USP siRNAs and USP19 siRNA were purchased from Horizon Discovery. USP19 adenovirus (ABM, Cat # 144298A) was purchased from ABM.

Cell Lines.

HEK293T (ATCC, Cat# CRL-3216, RRID: CVCL_0063), HeLa-GFP-TDP-43 (46), and HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Corning, 10-013-CV) supplemented with 10% fetal bovine serum (FBS, Sigma-Aldrich, 12,306C) and 1% penicillin/streptomycin (P/S, Gibco, 15,140-122).

Primary Neuronal Culture.

Primary mouse cortical and hippocampal neurons were prepared from P0 mice as previously described (72, 74). Briefly, both cortex and hippocampus were dissected in ice-cold HBSS buffer (Gibco, 14175-095) with 0.6% glucose, then digested with 0.25% trypsin-EDTA (Gibco, 25200056) at 37 °C. The tissues were homogenized and then gently centrifuged at 2,000 rpm, 4 °C, for 2 min. Cells were seeded on poly-D-lysine/laminin-coated cover glasses (Corning, 354087) or plates with neurobasal medium (Invitrogen, 10888022) containing 2% B27 (Invitrogen, 17,504,044), 2% GlutaMAX (Invitrogen, 35,050,061), and 1% penicillin/strep (Gibco, 15140-122).

Mice.

Usp19^+/− mice (75) were obtained from Simon Wing (McGill University, Canada). TAR4^+/− mice (56) were obtained from the Jackson Laboratory (Strain #: 012836; B6:SJL-Tg(Thy1-TARDBP)4Singh/J). Genomic DNA isolated from tail snips was used for genotyping by PCR as previously described (56, 75). All mice were maintained in the C57BL/6 background for at least three generations prior to breeding. All experimental procedures on mice were approved by the IACUC. Mice were housed together in an SPF facility with 2 to 4 littermates in sterile cages with pelleted bedding. Water and food were supplied ad libitum with 12-h light/dark cycle with mouse igloo enrichment. Individuals directly conducting studies on mice were blinded to their genotypes. All mice exhibited normal health without overt abnormalities until the end of experiment.

Reagents.

MG132 (Thermo Scientific, J63250-M), tetracycline (Gibco, Cat# A39246), tunicamycin (Millipore Sigma, Cat# 11089-65-9d), and thapsigargin (Millipore Sigma, Cat #: T9033) were purchased from the indicated sources.

Human Brain Tissue.

Paraformaldehyde-fixed floating frontal cortex tissues and frontal cortex brain chunks were obtained from the Alzheimer’s Disease Research Center at Emory University (Levey and Gearing). The primary neuropathologic diagnosis confirms AD and FTLD-tau patients (76). Primary pathologic diagnosis, Braak stage, ABC score, CERAD score, Thal score, age at onset, age at death, APOE genotypes, race, and sex are summarized in SI Appendix, Fig. S6.

DNA and siRNA Transfections and Viral Transductions.

DNA plasmids or siRNA (100 nM) were transiently transfected in HEK293T, HeLa-GFP-TDP-43, or HeLa cells with Lipofectamine 3000 (#3000015, Thermofisher) or Lipofectamine RNAiMAX (Cat #13778150, Thermofisher) in Opti-MEM I (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. The cells were harvested 48 h posttransfection. Control and USP19 adenovirus were used to transduce primary neurons.

Protein Extraction and Western Blotting.

The RIPA lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 2 mM ethylenediamine tetraacetic acid, 1% Triton X-100, 0.1% sodium dodecyl sulfate, 1% protease inhibitors mix (GeneDEPOT, P3200-005) and phosphatase inhibitors (GeneDEPOT, P3200-005) was applied to lyse brain tissues and cultured cells as previously described (72). RIPA lysates were centrifuged at 16,000 g for 15 min at 4 °C, and supernatants were transferred to new tubes (RIPA-soluble). The pellet was washed with ice-cold RIPA buffer and then lysed with 10% SDS buffer (10% SDS, 250 mM Tris pH 6.8) and sonicated in a water bath sonicator until no particulate material was visible (RIPA-insoluble). The Bicinchoninic Acid (Pierce™ BCA protein assay kit, 23225) method was applied to measure protein concentrations. Equal amounts of protein were subjected to SDS-PAGE and transferred to a nitrocellulose membrane (GE Healthcare, 10600002). Next, the membrane was blocked with 5% skim milk for 1 h at room temperature and next incubated with specified primary antibodies overnight at 4 °C. The peroxidase-conjugated secondary antibody was used for detecting the proteins by ECL (Thermo Scientific, 34580). The Fuji LAS-4000 imager (LAS-4000, Pittsburgh, PA, USA) or Amersham ImageQuant 800 (Cytiva) were used for imaging, followed by quantification using Image J software (NIH, https://imagej.nih.gov/ij/index.html).

Cytoplasmic and nuclear protein extraction were performed with NE-PER™ Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific, Cat # 78835), the extracted proteins were subjected to Western blotting to detect the corresponding protein level.

Primary antibodies used in Western blotting: Ubiquitin-K48 (Abcam Cat# ab140601), Ubiquitin-K63 (Abcam, Cat# ab179434), Myc (Cell signaling, #2276s), Flag (MA5-48063), Actin (Santa Cruz, sc-47778), TDP-43 (Cell signaling, #3448), Tubulin (Cell signaling, #2146); USP19 (Abcam, #93159), NUP98 (cell signaling, #2598s), ATF4 (Cell signaling, #11815s), ATF6 (Cell signaling, #65880S), CHOP (Cell signaling, #2895s), IRE1α (Cell signaling, #3294s), HA (Cell signaling, 2367s).

Proximity Ligation Assay (PLA) and Immunofluorescence.

USP19–TDP-43 protein complexes were detected in situ by PLA similar to previously described (47, 77, 78). Cells were washed with 1X PBS and fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature, blocked with 3% Normal Goat Serum (NGS) with 0.2% Triton X-100 at room temperature for 1 h, and incubated overnight with indicated primary antibodies at 4 °C followed by secondary antibodies (Alexa Fluor 488, Alexa Fluor 594) and DAPI (Thermo Scientific, 62248) or the Duolink® In Situ PLA® secondary antibody probes (Sigma-Aldrich). Negative controls excluded one primary or a secondary antibody probe. Primary antibody used: Ubiquitin (Cell signaling, #3936); Flag (ThermoFisher, MA5-48063), myc (Cell signaling, #2276s), USP19 (Abcam, #93159), TDP-43 (Proteintech, #67345), and Calnexin (Cell signaling, #2679).

Ex Vivo Slice Recordings.

Acute slice preparation from the hippocampus of 10-mo-old WT, TAR4^+/−, TAR4^+/−;usp19^+/−, and usp19^+/− mice was performed for electrophysiological recordings as we previously described (55, 72, 79). Briefly, a stimulating electrode was positioned on the Schaffer collaterals from the CA3 region. A glass recording electrode (1 to 4 MΩ) was placed in the hippocampal CA1 region. The electric stimulation, controlled by pClamp 11 software (Molecular Devices), was delivered via a stimulus isolator (model 2200; A-M Systems). The evoked field potentials were amplified using a differential amplifier (model 1800; A-M Systems), filtered at 1 kHz, and digitized at 10 kHz. In the CA1 region of a hippocampal slice, I-O curves were constructed from fEPSP, by stepping up stimulation amplitude which elicited half-maximal fEPSP from 1 to 15 mV at the rate of 0.05 Hz. PPF was evoked by two pulses with interpulse intervals from 20 to 300 ms. The percentage of the synaptic facilitation was measured by dividing the fEPSP slope elicited by the second pulse with the fEPSP slope elicited by the first pulse. LTP was induced by theta-burst stimulation and sampled for 60 min after the induction as described (72). LTP was estimated by dividing the slope of 60 min postinduction responses with the average slope of 20 min baseline responses.

Mouse Brain IHC and Human Brain IHC.

Please see SI Appendix, Methods.

Rotarod Testing.

Please see SI Appendix, Methods.

Statistical Analysis.

Graphs were created and analyzed by GraphPad Prism 10 software (GraphPad Software, San Diego, CA, USA). Group differences were analyzed by the t test (one-sample t test, or unpaired Student’s t test) ANOVA (one-way or, two-way ANOVA) followed by corresponding post hoc tests. All graphs are expressed as mean ± SEM.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix.