Caspase-4 mediates cytoplasmic accumulation of TDP-43 in the primate brains

The human mutant TDP-43(M337V) cDNA [51, 59] was subcloned into pAAV9-MCS or pGEX-4T1 vector (CellBiolabs) to generate AAV-mut-TDP-43(M337V) or GST-mut-TDP-43(M337V) vectors. The control AAV-GFP vector consisted of the same vector as for AAV-TDP-43 and contained the same ubiquitin promoter. Human caspase-4 or mouse caspase-11 cDNAs were generated using human or mouse brain tissues RNAs via RT-PCR, and cloned into pEGFP-C3, DesRed-C1, or AAV vector. AAV viruses (type-9) were packaged and amplified by the Viral Vector Core at Emory University. The titers of AAV vector genome were 1.4 × 10¹³ vg/ml for AAV-hTDP-43, 2.3 × 10¹³ vg/ml for AAV-GFP, and 1.8 × 10¹³ vg/ml for AAV-caspase-4. The following antibodies were used: mouse anti-human TDP-43 (Abnova, clone 2E2-D3) for the N-terminal TDP-43 (1-261 amino acids), rabbit anti-TDP-43 (Cell Signaling, G400) for a synthetic peptide corresponding to residues surrounding Gly400 of human TDP-43, rabbit anti-phospho-TDP-43 (Ser409/Ser410) (MABN14, Millipore), rabbit anti-LSD1 (2139S, Cell Signaling), rabbit anti-COX-IV (ab16056, Abcam), rabbit anti-PDI (2446S, Cell Signaling), rabbit anti-HSP90 (4874S, Cell Signaling), mouse anti-GFP (GTX628528, GeneTex), rabbit anti-NeuN (ABN78, Millipore), rabbit anti-GFAP (AB5804, Millipore), rabbit anti-caspase-4 (4450, Cell Signaling) and mouse anti-GST (Santa Cruz, A-6). All secondary antibodies were purchased from Jackson Immuno-Research Laboratories. Inhibitors of caspase family members included ZVAD-fmk (Selleckchem), DEVD-fmk (Tocris Bioscience), LEHD-fmk (Tocris Bioscience), and LEVD-fmk (BioVision).

Monkeys were housed in accordance with Chinese National standards, which are consistent with the standard set forth in the eighth edition of the NRC Guide for the Care and Use of Laboratory at the Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences, Peking Union Medical College, which is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC), International. The animal use and experiments followed the protocol that was approved by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences. This study was conducted in strict compliance with the “Guide for the Care and Use of Laboratory Animals of the Institute of Laboratory Animal Science (est. 2006)” and “The use of non-human primates in research of the Institute of Laboratory Animal Science (est. 2006)” to ensure the personnel safety and animal welfare.

All mice (C57BL/6) were bred and maintained in the animal facility at Emory University under specific pathogen-free conditions in accordance with institutional guidelines of the Animal Care and Use Committee at Emory University. The studies followed the protocol approved by the Animal Care and Use Committee at Emory University.

Of six AAV-TDP-43- and six AAV-GFP-injected male monkeys at age of 8–12 years, four AAV-TDP-43- and two AAV-GFP-injected monkeys were selected, based on their similar body weights and ages, for behavioral analysis once each week for 12 weeks.

The movement capabilities of monkeys were tested as described previously [49]. Briefly, we carefully observed behavior of the monkeys during the day time and recorded video monitoring for 30 min per week. To evaluate weakness of the forelimb muscles on the AAV-injected monkeys, we performed the “open-field test” with continuous video recording. The “apple test” was performed and placed in line from back (monkey side) to front (observer side), to estimate the spontaneous action. In the “fence or grasping test”, we analyzed how frequently a monkey used his/her left or right hand to grasp the ceiling fence or test rod by video recording 30 min for each session each day. Four sessions were analyzed for obtaining the percentage of using the left or right hand. To evaluate weakness of the upper limb muscles, we measured grip strength using a spring hand dynamometer connecting with a small handle made specially for measuring monkey upper limb strength (AiDebaoHandPink, China). In this test, a monkey was allowed to grasp the handle with one upper limb (left or right), and its muscle strength was measured by the examiner pulling the spring hand dynamometer until the animal released the handle. The digital score of the force on the grip strength meter was recorded for statistical analysis.

Human cortex tissues were obtained and archived via an institutional review board and Health Insurance Portability and Accountability Act compliant process at neuropathology/histochemistry core of Emory University. Autopsies occurred following death with a postmortem interval of 6–13 h. The cortex tissues were obtained from the postmortem brains of ALS patients (E04-56, E11-75, E09-35, E08-86 and E11-81) who died at 67–74 years of age and were confirmed to have TDP-43 aggregates via postmortem histologic analysis. Non-ALS control tissues were obtained from neurologically unaffected patients (E06-45, E06-114, E08-101, E08-137 and E10-142) who died at 53–92 years of age.

For analyzing TDP-43 expression and distribution in the monkey brain, we used three AAV-TDP-43 or AAV-GFP monkeys for Western blotting and another three AAV-TDP-43 or AAV-GFP monkeys for immunocytochemical studies. Animals were anesthetized and perfused with 10 ml 0.9% NaCl, and then with 20 ml of 4% paraformaldehyde in 0.1 M PBS through the left cardiac ventricle. Brains were removed and fixed overnight in the same solution and cryopreserved with 15% and 30% sucrose before sectioning into 10 µM sections with a cryostat (Leica CM1850) at − 20 °C. Sections from monkey or mouse brains or cultured cells were fixed in 4% paraformaldehyde in PBS for 10 min, permeabilized with 0.2% Triton X-100 in PBS for 30 min, blocked with 3% normal donkey serum in 3% BSA for 1 h, and incubated with primary antibodies in 3% BSA overnight at 4 °C. After several washes with PBS, the brain sections or fixed cells were incubated with secondary antibodies conjugated with either Alexa-488 or Alexa-594 (Invitrogen). 0.01 μg/ml DAPI was used to label the nuclei. Fluorescent images were taken with a Zeiss Axiovert 200 MOT microscope of the 40 ×/0.6 lens or 63 ×/0.75 lens, equipped with a digital camera (Hamamatsu, Orca-100) and Openlab software (Improvision). The immunostaining analysis of TDP-43 subcellular distribution in the injected monkey or mouse brains was performed completely blinded on standardized 40 mm sections. The monkey brain sections were prepared using a brain slicer including the injected regions (3 substantia nigra: pars compacta/SNpc, pars reticulate/SNpr and pars lateralis/SNpl). Each brain region was used to take at least six images (40 × magnification) that can clearly reveal the subcellular distribution of TDP-43. For the quantitative analysis of differential subcellular location of TDP-43 in the monkey and mouse brain, the numbers of cells showing the nuclear or cytoplasmic TDP-43 per image were presented as the mean ± SEM, and the quantitative data were obtained from three monkeys or six mice per group. Densitometry analyses of fluorescent intensities of aggregates were quantified by ImageJ software (W. Rasband, National Institutes of health, USA).

Monkey or mouse brain tissues were homogenized for 25 strokes with a dounce homogenizer ice-cold buffer (0.32 M sucrose, 15 mM Tris–HCl, 60 mM KCl, 15 mM NaCl, 5 mM EDTA, 1 mM EGTA, 0.02% NaN₃, 2 mM ATP, pH 8.0) containing protease inhibitor (Roche) and 100 μM PMSF. Ten percent lysates were stored as the total lysate sample. Nuclei and cellular debris were pelleted (P1) at 800 × g for 5 min. The supernatant (S1) was transferred to a new tube and centrifuged at 20,000 × g for 30 min at 4 °C to obtain the mitochondria-enriched pellet (P2). The supernatant (S2) was then used for the soluble cytoplasmic fraction. The S2 was centrifuged at 100,000 × g for 30 min at 4 °C to obtain the endoplasmic reticulum-enriched pellet (P3). Crude nuclear pellets were washed four times with ice-cold homogenization buffer to remove cytoplasmic contaminants. For nuclear purification, the pellets were re-suspended in 374 μl of buffer [15 mM HEPES, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM DTT, 26% glycerol (v/v), pH 7.9] with 26 μl of 4.6 M NaCl to generate the final concentration at 300 mM NaCl, homogenized with 20 full strokes in Teflon homogenizer on ice, and sonicated for 10 s. The homogenized samples were kept on ice for 20 min and then centrifuged at 24,000 × g for 20 min at 4 °C.

All the tissue samples were adjusted to 0.5 mg/ml total protein by dilution with homogenization buffer for triplicate caspase assays. Caspase-4 activity was determined using the specific Ac-YVAD-AFC substrate (10 μg/ml; BioVision). Equal amounts (10 μg) of the extracts were incubated with corresponding substrates in 100 μl caspase-4 activity assay buffer (0.05 M Tris–HCl, 0.5 mM EDTA, 1 mM ATP, 1 mM DTT, pH8.0) for 30–60 min at 37 °C. Cold water (0.8 ml) was used to stop the reactions, and the reaction mixtures were iced for at least 10 min. The free AFC fluorescence was quantified using the CytoFluor multi-well plate reader (FLUOstar; BMG LABTECH) with excitation and emission wavelengths at 400 nm and 500 nm, respectively. All readings were standardized using the fluorescence intensity of an equal volume of free 7-amino-4-trifluoromethylcoumarin (AFC) solution (40 mM), normalized by the protein concentrations and expressed as nmol/min/mg protein.

Purified GST-mut-TDP-43 (M337V) in Sepharose beads were diluted in cold assay buffer (25 mM Tris–HCl, 10 mM MgCl₂, 100 µg/ml purified rabbit creatine kinase, 50 mM phosphocreatine, 1 mM ATP, pH 7.6). Monkey or mouse tissues form the brain striatum and cortex were homogenized at 1 g/ml in cold assay buffer using 20 strokes of a glass dounce hand homogenizer and were centrifuged at 500 × g for 5 min at 4 °C to pellet unbroken tissues and membranes. The supernatant was collected and stored on ice, while protein concentrations were determined using a BCA Protein Assay Kit (Thermo Scientific). The lysates (200 μl) at 0.5 mg protein/ml were incubated with GST-TDP-43 beads (200 μl) at 37 °C with 300 rpm shaking for 24 h. The beads were centrifuged and combined with the protein loading dye (0.2% SDS) for SDS-PAGE and Western blotting analysis to detect expression of the cleaved GST-TDP-43 using an anti-GST antibody. An anti-C-terminal TDP-43 antibody was used to detect the presence of C-terminal TDP-43 in the supernatant. Different caspase inhibitors at the concentrations (50–100 μM) were included in the lysates to inhibit caspase activity during the incubation process with GST-TDP-43 beads.

Cultured cells were transfected with plasmid DNAs using RNAi Max transfection reagent (Invitrogen) according to the manufacturer’s protocol. At 48 h following transfection, cells were harvested for immunofluorescent staining and Western blotting. For caspase-4 knockdown experiment, the cultured human neural SH-SY5Y cells were transiently transfected with caspase-4 siRNA (Gene Pham Co. sequence GUGUAGAUGUAGAAGAGAAtt or AAGUGGCCUCUUCACAGUCAUtt) or control siRNA (scrambled sequence) using RNAi Max transfection reagent (Invitrogen) according to the manufacturer’s protocol. At 48 h following transfection, cells were harvested for immunofluorescent staining and Western blotting.

The cell viability assay was done using Cell Counting Kit-8 (CCK-8) (Dojindo, Japan), which determines the number of viable cells in proliferation and cytotoxicity. It utilizes tetrazolium-8-[2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium] monosodium salt that produces a water-soluble formazan dye in living cells. The amount of this dye is directly proportional to the number of living cells, offering a sensitive assay to detect viability of many cell lines including SH-SY5Y cells [10, 56]. Briefly, SH-SY5Y cells were transfected with siRNA for 48 h and then treated with tunicamycin (1 μg/ml) for 12 h. An equal number of 100 µl SH-SY5Y cell suspension (1 × 10⁴ cells/100 µl/well) was dispensed in a 96-well plate. Each well of the plate was added with 10 µl CCK-8 solution, and incubated for 2 h at 37 °C. The absorbance values at the wavelength of 450 nm were spectrophotometrically measured using an CytoFluor multi-well plate reader (FLUOstar; BMG LABTECH). Each sample was tested in three replicates. SH-SY5Y cells without transfection and tunicamycin treatment were used as negative and blank controls. The values (mean ± SEM) of treated SH-SY5Y cells were calculated as  % of control cells.

Our previous studies demonstrated that stereotaxic injection of mutant TDP-43 in the striatum of mice led to abundant nuclear TDP-43 in the injected brain [59]. Using an adenoviral-associated vector (AAV) that expresses human TDP-43 (M337V) under the control of the ubiquitin promoter (Fig. 1a) and the same stereotaxic approaches in our previous study [60], we injected AAV-TDP-43 into the right substantia nigra of six male rhesus monkeys (8–12 years) and twelve mice (10–14 months). In our experiments, we used the substantia nigra in the monkey because this small brain region is suitable for viral injection and controls the movements of the limbs and elbows [31, 54], allowing us to examine the subcellular distribution of TDP-43 as well as the phenotypes associated with TDP-43 toxicity. Upper limb and elbow movements were measured to determine the effects in the animals. The controls included the injection of the same AAV vector-expressing GFP (AAV-GFP).

Three months after the injection, the AAV-GFP monkeys did not show any motor symptoms, but all the TDP-43-injected monkeys developed the obvious weakness of the left upper limb, whose movement is controlled by the contralateral right substantia nigra that was injected with AAV-TDP-43 (Fig. 1b). These upper limb weaknesses appeared 2–4 weeks post injection, increased in severity, and stabilized after 3–4 months. Daily video showed that TDP-43-injected monkeys displayed significantly infrequent use of their left upper limbs as compared with the GFP-injected control monkeys when walking and grasping food, the ceiling fence, or a rod (Suppl. Figure 1; Suppl. Video.1). Because these monkeys could not raise left upper limbs and straighten left elbow joints, they had difficulties reaching and grasping the celling fence or a test rod (Fig. 1c). Some monkeys had paralyzed upper limbs, resulting in walking difficulties. By measuring the grip strength using a spring hand dynamometer, we found that TDP-43-injected monkeys had a significant decrease in the left upper limb muscle strength compared with the right upper limb (Fig. 1d). The severe limb movement impairment could be due to the acute and broad toxicity of overexpressed TDP-43 in the injected substantia nigra. However, no early death occurred in all monkeys injected with AAV-TDP-43 (Fig. 1e). Similarly, overexpressing TDP-43 (M337V) in the right substantia nigra in mice also led to markedly reduced grip strength of the left limb (Fig. 1f, g) and poor rotarod and balance beam performance within 2 weeks after viral injection (Suppl. Figure 2).

The neurological phenotypes of the TDP-43-injected mice and monkeys confirmed the neurotoxicity of overexpressed mutant TDP-43 and also allowed us to compare TDP-43’s expression and localization in these two different species. The brains of mice and monkeys were dissected after euthanization and prepared for Western blotting and immunocytochemistry. We used an antibody against a C-terminal peptide corresponding to residues surrounding Gly400 of human TDP-43 to detect the overexpressed exogenous TDP-43. Immunocytochemistry revealed that injection of the control AAV-GFP yielded diffuse green fluorescence throughout the entire cell (Suppl. Figure 3). However, the injected mutant TDP-43 was diffusely distributed in the cytoplasm or adjacent to the nuclei in the monkey brain (Fig. 2a). In contrast, the injected mouse brain showed the predominant nuclear distribution of mutant TDP-43 (Fig. 2a), similar to the abundant nuclear localization of transgenic TDP-43 found in different mouse models [14, 19, 43, 52, 59]. Quantification of the percentage of cells showing cytoplasmic or nuclear accumulation of transgenic TDP-43 in the injected brain region also verified the remarkable differences in the subcellular distribution of mutant TDP-43 in the mouse and monkey brains (Fig. 2a).

The above findings have shown that different subcellular localization of TDP-43 is species-dependent even when mutant TDP-43 is overexpressed. To investigate whether there is any difference in TDP-43 cleavage, we performed Western blotting using anti-C-terminal TDP-43 antibody. Interestingly, truncated TDP-43 fragments (35- and 25-kD) were abundant in the monkey brain, whereas full-length TDP-43 is expressed at a comparable level as endogenous TDP-43 in the AAV-GFP-injected monkey. However, in the mouse brain, full-length TDP-43 was much more abundant than C-terminal truncated TDP-43, which is consistent with the previous findings from transgenic TDP-43 mice [34, 52, 55]. The low level of truncated TDP-43 in the mouse brain is remarkably different from the abundant level of truncated TDP-43 in the monkey brain (Fig. 2b). The quantification of the relative level of intact and truncated TDP-43 on Western blotting also verified significant differences in TDP-43 cleavage in monkey and mouse brains (Fig. 2c). Western blotting revealed a significant reduction of the neuronal marker NeuN and an increase in GFAP staining, which reflects neuronal toxicity and reactive astrocytes, in both mouse and monkey brains (Fig. 2d, e). Double immunostaining verified the reduced number of NeuN-positive cells in AAV-TDP-43 injected monkey brain region (Suppl. Figure 4), which was similar to what we found in the AAV-TDP-43 injected mouse brain [59]. These alternations indicate neurodegeneration in the injected brain regions, and are consistent with the phenotypes of the injected monkeys and mice. Using an antibody to N-terminal TDP-43 (1-261 amino acids) or phosphorylated S409/S410 TDP-43, we also found that TDP-43 is readily cleaved in the monkey brain, but not in the mouse brain (Suppl. Figure 5). These findings suggest that monkey brain tissues can selectively cleave TDP-43 to generate C-terminal TDP-43 fragments, which may accumulate in the cytoplasm.

The above striking difference led us to examine the differential processing of TDP-43 in the monkey and mouse brains. TDP-43 consists of 414 amino acids with NLS in the N-terminal region and NES in the region close to the C-terminus (Fig. 3a, Suppl. Figure 6). We generated PRK vectors to express N-terminal fragments (1–173), truncated (77–414), C-terminal (174–414), or full-length TDP-43 under the CMV promoter (Suppl. Figure 6). The expression of these TDP-43 proteins in transfected N2A cells was detected using Western blotting and immunostaining with antibodies that recognized C-terminal or N-terminal TDP-43. Both truncated (77–414) and C-terminal (174–414) fragments of TDP-43 were distributed in the cytoplasmic region. In contrast, N-terminal (1-173) TDP-43 with the NLS signal sequences is localized in the nucleus (Suppl. Figure 7). These results suggest that the specific processing of TDP-43 in the monkey brains may generate TDP-43 fragments that lack N-terminal NLS and are able to accumulate in the cytoplasm.

To test the above idea, we generated recombinant GST fusion proteins containing mutant TDP-43 (M337V) that was produced in bacteria and then purified (Fig. 3a). Purified GST-TDP-43 proteins in glutathione-affinity beads were incubated overnight with the monkey or mouse brain lysates at 37 °C for in vitro cleavage. Because of the limited substantia nigra tissues, we used the striatum and cortex tissue lysates from monkeys and mice in this in vitro cleavage assay. The beads were precipitated and immuno-blotted by anti-GST to detect cleaved GST-TDP-43 products (Fig. 3b). This in vitro analysis revealed that monkey brain tissues could cleave full-length TDP-43 to generate fragments (32- and 42-kD) that were detected by anti-GST and anti-N-terminal TDP-43 (Fig. 3c). Generation of these GST fusion protein fragments corresponds to the production of cleaved C-terminal TDP-43 products (25- and 35-kD) (Fig. 3a), which are present in the monkey brain (Fig. 2b, c) and in the in vitro cleavage assay (Suppl. Figure 8). Using different subcellular fractions from monkey and mouse brain tissues to incubate with full-length GST-TDP-43 (Suppl. Figure 9), we found that the cytosolic fraction had the greatest activity to generate fragmented TDP-43 than the nuclear and mitochondrial fractions (Fig. 3d), suggesting that TDP-43 is cleaved by cytosolic molecules in the monkey brains. We then applied MG132 and BFA to inhibit the ubiquitin–proteasome system and autophagy, respectively, and found that these inhibitors did not abolish the generation of TDP-43 fragments but could increase the levels of fragmented TDP-43 produced by monkey brain lysates (Suppl. Figure 10a). These findings suggest that a specific cytosolic enzyme rather than the UPS and autophagy in the monkey brain may generate TDP-43 fragments, which are then degraded by the UPS or autophagy.

In vitro experiments revealed that multiple caspases (caspases-3, -4, -7 and -9) cleave TDP-43 in cultured cells [6, 11, 26, 29, 42]. However, these in vitro experiments introduced caspases by transfection into cultured cells and did not tell us whether some of these caspases can endogenously determine the specific cleavage of TDP-43 in the primate brains. To investigate whether caspases generate the specific cleavage products of TDP-43 in the monkey brain, we first used the pan-caspase inhibitor, ZVAD-fmk to inhibit all the caspase family members and found that this inhibitor completely blocked the generation of fragmented TDP-43 by the monkey brain lysates (Fig. 3e). Since the in vitro study demonstrated that TDP-43 was initially cleaved after Asp174 by caspase-4 [29], we added the caspase-4-specific inhibitor LEVD-fmk and found that it could effectively block the cleavage of TDP-43 and prevent the generation of 42- and 32-kD fragments (Fig. 3f). However, neither of DEVD-fmk inhibitors of caspase-3 and -7, nor LEHD-fmk inhibitor of caspase-9 could prevent the initial cleavage of TDP-43 (Suppl. Figure 10b, c).

Caspase-4 is an enzyme that cleaves proteins proteolytically at an aspartic acid residue and contributes to inflammatory activation in mono-myelocytic cells [4, 44, 50]. Interestingly, by analyzing amino acid sequences of human caspase-4 and its homologues in monkey and mouse, we found that caspase-4 is only present in primates and its rodent homologue is caspase-11. Caspase-4 in humans and monkeys share up to 95% identify in amino acid sequences, but human caspase-4 shares only 50% identify with mouse caspase-11 (Suppl. Figure 11). Immuno-blotting with the antibody to human caspase-4 showed that caspase-4 is widely present in different brain regions and peripheral tissues in monkeys, while the anti-caspase-4 failed to detect any band in the mouse tissues (Fig. 4a).

Caspase-4 is bound to the endoplasmic reticulum (ER) and can be activated under ER stress [18]. Previous reports note that overexpression of TDP-43 or other misfolded proteins induces ER stress [32, 36, 37]. A fluorescence assay using caspase-4 substrate Ac-YVAD-AFC indeed revealed elevated caspase-4 activity in TDP-43-injected monkey brain as compared with GFP-injected monkey (Fig. 4b). The cleaved and activated caspase-4 (between 25- and 45-kD) were also up-regulated in the TDP-43-injected monkey substantia nigra as compared with the control GFP-injected monkey (Fig. 4c). This finding explains the data in Fig. 2B that shows more abundant cleaved TDP-43 fragments than full-length mutant TDP-43 in the TDP-43-injected monkey brain. Further, an in vitro cleavage assay showed the increased generation of cleaved TDP-43 products by TDP-43-injected monkey brain lysates compared with GFP-injected monkey brain, and this increase could be completely inhibited by the caspase-4 specific inhibitor LEVD-fmk (Fig. 4d).

A previous study reported the increased amounts of caspase-4 in the spinal cords of ALS patients [3]. Western blotting analysis of the postmortem brain cortex tissues from 5 ALS patients and 5 non-ALS individuals revealed that the activated form (between 25- and 45-kD) of caspase-4 was indeed increased in the ALS samples (Fig. 5a, Suppl. Table 1). Quantitative analysis of the ratios of the activated caspase-4 to the full-length caspase-4 or the loading control GAPDH on the same blots also confirmed the selective increase of activated caspase-4 in the ALS patient brains (Fig. 5b). The increased caspase-4 activity in postmortem ALS patient brain tissues allowed us to explore whether this increased activity is associated with the generation of TDP-43 fragments. Thus, we incubated GST-TDP-43 with postmortem brain tissue lysates. Consistently, lysates from ALS patient brains also produced more TDP-43 fragments than the brain lysates from control individuals (Fig. 5c). To verify that TDP-43 is selectively cleaved by caspase-4 but not mouse homologous caspase-11, we co-transfected pEGFP-TDP-43 (M377V) with DesRed–human caspase-4 or mouse caspase-11 in mouse N2A cell line. The green fluorescent labeling shows the cytoplasmic localization of transfected TDP-43 in N2A cells when co-transfected with human caspase-4 (red). However, mouse caspase-11 transfection was unable to cause the cytoplasmic distribution of TDP-43 (Fig. 5d, e) and to generate of the truncated TDP-43 (Fig. 5f), supporting the idea that only caspase-4 cleavage generates TDP-43 fragments that are able to accumulate in the cytoplasm.

If caspase-4 is absent in the mouse brain, but is responsible for the cytoplasmic accumulation of TDP-43, transgenic expression of caspase-4 in the mouse brain should also lead to the cytoplasmic accumulation of TDP-43. We generated AAV vector-expressing human caspase-4 under the control of the ubiquitin promoter (Fig. 6a) and confirmed its expression in cultured N2A cells (Fig. 6b). We then stereotaxically delivered this virus with AAV-TDP-43 (M337V) into the substantia nigra in 10–14 months old mice. As expected, immunofluorescent staining showed abundant nuclear aggregates of transgenic TDP-43 in the mouse brain in the absence of caspase-4. However, co-expression of caspase-4 caused mutant TDP-43 to distribute in the cytoplasm (Fig. 6c). Furthermore, Western blotting with the antibody to C-terminal TDP-43 confirmed that co-expression of caspase-4 could yield C-terminal 25-kD and 35-kD TDP-43 fragments (Fig. 6d, e).

To provide direct evidence for the causative role of casapse-4 in the cleavage of TDP-43 in cells, we used siRNAs to knockdown the expression of endogenous caspase-4 in cultured human neural SH-SY5Y cells. The siRNA (siCAS4-1) had higher efficiency on silencing endogenous caspase-4 than siCAS4-2 (Suppl. Figure 12). The siCAS4-1 was then transfected into SH-SY5Y cells for 48 h, and the transfected cells were treated with tunicamycin (1 μg/ml) for 12 h, which is a known ER stress activator for caspase-4 induction [18, 58]. As expected, immunofluorescent staining showed abundant cytoplasmic distribution of endogenous TDP-43 after tunicamycin treatment. Moreover, this cytoplasmic distribution was reduced by suppressing caspase-4 expression via siRNA (siCAS4-1) (Fig. 7a). Western blotting also showed that suppressing the expression of endogenous caspase-4 could significantly reduce the generation of C-terminal 25-kD and 35-kD TDP-43 fragments (Fig. 7b, c). Tunicamycin has been reported to increase caspase-4 activity [18, 58] and reduce the viability of SH-SY5Y cells [16, 61]. We therefore measured cell viability of the cultured SH-SY5Y cells using CCK-8, a widely used assay that can sensitively measure the viability of a variety of cultured cells including SH-SY5Y cells [10, 56]. We found that inhibition of caspase-4 via siRNA-caspase-4 also significantly increased cell viability after tunicamycin treatment in parallel with reduction of cytoplasmic accumulation of endogenous TDP-43 (Fig. 7d).

If TDP-43 is cleaved by activated caspase-4, increased activity or expression of caspase-4 should promote the cytoplasmic distribution of endogenous TDP-43 in the monkey brain. To test this idea, we injected the same titer of AAV-caspase-4 or control AAV-GFP into the prefrontal cortex in 11-year-old wildtype monkeys. After 3 weeks, the injected brain prefrontal cortex was isolated for immunofluorescent staining with antibodies to caspase-4 and C-terminal TDP-43. Double immunofluorescent staining showed that endogenous TDP-43 in the normal monkey brain is predominantly distributed in the nucleus (Fig. 8a). However, in the AAV-caspase-4 injected monkey brain region, endogenous TDP-43 shows the cytoplasmic and diffuse distribution neuronal cells (Fig. 8b, c). This cytoplasmic distribution is unlikely due to stereotaxic injection and overexpression of AAV vector, as quantitative data show that injection of AAV-GFP was unable to alter the distribution of endogenous TDP-43 in the nucleus (Fig. 8a, c). Because caspase-4 is expressed in the primates but not rodents, we propose that caspase-4 mediates the unique cytoplasmic accumulation of mutant TDP-43 in primates including humans (Fig. 9).