Elimination of TDP-43 inclusions linked to amyotrophic lateral sclerosis by a misfolding-specific intrabody with dual proteolytic signals

Aggregation of TAR DNA-binding protein of 43 kDa (TDP-43) is implicated in the pathogenesis of sporadic and certain familial forms of amyotrophic lateral sclerosis (ALS), suggesting elimination of TDP-43 aggregates as a possible therapeutic strategy. Here we generated and investigated a single-chain variable fragment (scFv) derived from the 3B12A monoclonal antibody (MAb) that recognises D247 of the TDP-43 nuclear export signal, an epitope masked in the physiological state. In transfected HEK293A cells, 3B12A scFv recapitulated the affinity of the full-length MAb to mislocalised TDP-43 with a defective nuclear localising signal and to a TDP-43 inclusion mimic with cysteine-to-serine substitution at RRM1. Moreover, 3B12A scFv accelerated proteasome-mediated degradation of aggregated TDP-43, likely due to an endogenous PEST-like proteolytic signal sequence in the VH domain CDR2 region. Addition of the chaperone-mediated autophagy (CMA)-related signal to 3B12A scFv induced HSP70 transcription, further enhancing TDP-43 aggregate clearance and cell viability. The 3B12A scFv also reduced TDP-43 aggregates in embryonic mouse brain following in utero electroporation while causing no overt postnatal brain pathology or developmental anomalies. These results suggest that a misfolding-specific intrabody prone to synergistic proteolysis by proteasomal and autophagic pathways is a promising strategy for mitigation of TDP-43 proteinopathy in ALS.

SCieNtifiC RepoRTS | (2018) 8:6030 | DOI: 10.1038/s41598-018-24463- 3 We next compared the specific reactivities of 3B12A intrabodies to TDP-43 species using sandwich ELISA assay. Lysates from cultured HEK293A cells co-transfected with Myc-tagged 3B12A intrabodies and FLAG-tagged TDP-43 or SOD1 species were incubated on ELISA plates previously coated with anti-FLAG antibody and binding subsequently detected with anti-Myc antibody (Fig. 2e,f) or anti-TDP-43 antibody ( Supplementary Fig. S4a). Quantitative analysis by sandwich ELISA showed the preferential interaction of VH_VL scFv with the cytoplasmic aggregation-prone TDP-43 mutant TDP-43 mNLS,C173S/C175S compared with VL_VH, VH and VL, as well as the lower affinity to TDP-43 WT and SOD1 species including SOD1 G93A mutant (Fig. 2e). The VH_VL intrabody also showed higher affinity to familial ALS-linked A315T and Q331K mutants of TDP-43 than to TDP-43 WT (Fig. 2f) although affinity was lower than to the C173S/C175S mutant. Moreover, both 3B12A scFv forms showed a higher affinity to TDP-43 mNLS,C173S/C175S than C4F6 scFv, an intrabody derived from C4F6 MAb 35 raised against misfolded SOD1 mutant SOD1 G93A (Supplementary Fig. S4b). Thus, all four 3B12A intrabodies demonstrated high affinity for misfolded TDP-43, particulally VH_VL scFv, although reactivity to the native form varied. Furthermore, VH_VL scFv also showed a higher reactivity to other ALS-linked TDP-43 mutants than to WT. However, the discrepancy between sandwich ELISA and immunoprecipitation experiments regarding aggregate binding of VH or VL nanobodies implies that a nonspecific interaction, possibly based on the insolubility of nanobodies, may be responsible for the interactions of these forms with aggregated TDP-43 observed in ELISA. Based on these results, we used VH_VL scFv for subsequent examination of TDP-43 clearance.
Chaperone-mediated autophagy signal confers bi-directional proteolytic properties to 3B12A VH_VL scFv. Self-degradation is a desirable property for a therapeutic intrabody, so we investigated the effects of the VH domain PEST-like sequence and protein disassembling signals on the degradation of 3B12A VH_VL scFv via proteasomal and autophagic-lysosomal proteolysis pathways using a cycloheximide (CHX) chase study. For this purpose, we constructed a VH_VL intrabody containing two protein disassembling signals, CL1 for proteasome localisation 36,37 and CMA for chaperone-mediated autophagy 38,39 (Fig. 3a), based on the assumption that the interaction between scFv and antigen alone may not suffice for protein clearance. The chase assay revealed that 3B12A scFv (with a PEST-like sequence) declined more rapidly than C4F6 scFv without a PEST-like sequence, and proteasome inhibition by lactacystin prevented this decline (Fig. 3b,c). The intracellular concentration of modified 3B12A scFv containing the CL1 signal (VH_VL-CL1) also declined rapidly over 10 h following CHX application, and this decline was prevented by lactacystin (Fig. 3b,c). Furthermore, VH_VL-CL1 detection by Western blot analysis required a longer ECL exposure time than required for VH_VL alone, suggesting that the CL1 signal accelerated VH_VL degradation in proteasomes. The intracellular level of 3B12A scFv containing a CMA signal (VH_VL-CMA) declined as rapidly as VH_VL and VH_VL-CL1, and this decrease was  inhibited by the lysosome inhibitor bafilomycin as well as by lactacystin (Fig. 3b,c). Consistent with lysosomal degradation, the effect of bafilomycin was significantly greater on CMA-fused VH_VL than VH_VL-CL1 or VH_ VL alone. Lactacystin significantly reversed the degradation of all 3B12A scFv types. The lack of a significant difference in elimination rate among these different VH_VL forms may be due to the endogenous PEST-like signal. Indeed, less of the VH fragment (containing the PEST-like sequence) remain at 10 h after the chase compared with the VL fragment, and lactacystin significantly inhibited VH degradation. This finding indicates that the VH domain PEST-like sequence regulates its turnover (Fig. 3d,e). We confirmed that neither CHX, lactacystin, (c) Quantitative analysis of (b) Each data point was obtained by normalisation to actin. Differences were evaluated by one-way ANOVA (mean ± SD from three independent experiments; *p < 0.05, ***p < 0.005 and ****p < 0.001). N.S. indicates not significant. 'remaining scFv (%)' indicates 'the % signal compared with time 0' . (d) Protein degradation chase assay of 3B12A-VH and 3B12A-VL in HEK293A cells. (e) Quantified analysis of (d) Each data point was obtained by normalisation to actin. Differences were evaluated by one-way ANOVA (mean ± SD from three independent experiments; **p < 0.01 and ***p < 0.005). nor bafilomycin was cytotoxic, further indicating that intrabody decline resulted from specific proteasome-and lysosome-mediated degradation ( Supplementary Fig. S5).
We also confirmed that proteasome activity was not suppressed by 3B12A scFv alone ( Supplementary  Fig. S6a), TDP-43 WT or misfolded TDP-43 species (Supplementary Fig. S6b) in the presence and absence of 3B12A scFv ( Supplementary Fig. S6c). Likewise, measurement of LC3-II/LC3-I flux by Western blot analysis showed that autophagy-lysosomal activity was not altered by intrabodies containing protein disassembling signals in the presence and absence of WT or misfolded TDP-43 ( Supplementary Fig. S6d,e).
3B12A scFv-CMA facilitates the degradation and elimination of misfolded TDP-43 in the cytosol. To evaluate the efficacy of the 3B12A intrabody for intracellular clearance of misfolded TDP-43, we performed HaloTag pulse chase assays in which a covalent bond between a HaloTag-fused protein and synthetic ligands enables the transient labelling of target proteins in live cells. HEK293A cells overexpressing HaloTag-fused TDP-43 with or without VH_VL 3B12A scFv intrabody co-expression were labelled with a cell membrane-permeable HaloTag fluorescent ligand (HaloTag diAcFAM ligand), and cell lysates were analysed at several time points post-labelling. The HaloTag pulse chase assay demonstrated that 3B12A scFv co-expression markedly enhanced the degradation of the cytoplasmic aggregation-prone TDP-43 mutant TDP-43 mNLS,C173S/C175S at 12 and 24 h after adding HaloTag diAcFAM ligand, whereas 3B12A scFv did not promote clearance of TDP-43 WT (Fig. 4a-d). Moreover, the addition of a CMA signal (VH_VL-CMA) accelerated the degradation of aggregated TDP-43 compared with VH_VL 3B12A scFv or VH_VL-CL1 (Fig. 4b,d). Co-expression of VH_VL-CMA also significantly enhanced the degradation of the nuclear aggregated TDP-43 mutant (TDP-43 C173S/C175S ) but did not significantly accelerate clearance of the cytoplasmic TDP-43 mutant (TDP-43 mNLS ) (Fig. 4e,f), confirming that VH_VL-CMA has higher affinity to misfolded species. Furthermore, this effect was inhibited by either lactacystin or bafilomycin (Fig. 4g,h), indicating that cytoplasmic TDP-43 aggregates are eliminated by VH_VL-CMA through both the proteasomal and autophagy-lysosome pathways. We also conducted biochemical analysis of cytotoxicity and cell death in the neuroblastoma cell line N2a to evaluate whether 3B12A scFv can rescue neuronal cells as well as non-neuronal HEK292A cells. Neuroprotection against TDP-43 aggregates by 3B12A scFv was evaluated by simultaneous treatment of transfected N2a cells with membrane-impermeable bis-AAF-R110 and cell-permeable GF-AFC, fluorescent compounds that react with proteases in dead and live cells, respectively. Among the TDP-43 species tested, cytoplasmic TDP-43 aggregates (TDP-43 mNLS,C173S/C175S ) induced the highest cytotoxicity, consistent with our previous studies in the NSC-34 cell line 33 . This cytotoxicity was significantly reduced in the presence of 3B12A scFv (VH_VL), and the inclusion of the CMA signal significantly enhanced this protective effect ( Fig. 5e-g), while 3B12A scFv species alone (vector controls) did not induce neurotoxicity or otherwise affect N2a cell viability.
3B12A scFv-CMA induces HSP70 in the presence of misfolded TDP-43 aggregates. The intrabody 3B12A scFv-CMA exhibited the high efficiency for clearance of cytoplasmic TDP-43 aggregates (Fig. 4d) and for reducing neurotoxicity induced by TDP-43 aggregates (Fig. 5f). The CMA is a selective proteolytic pathway associated with the heat shock protein 70 (HSP70) family member heat shock cognate protein 70 (HSC70) 39,40 , which functions as a molecular chaperone for the elimination of misfolded proteins. Therefore, we first examined if 3B12A scFv-CMA altered the expression of HSP70 family proteins including HSC70 and HSP70 in HEK293A cells.
Western blot analysis revealed that HSP70 was upregulated, but only when cells were cotransfected with 3B12A scFv-CMA and TDP-43 mNLS,C173S/C175S (Fig. 6a-c). In contrast, transfection had no effect on expression of HSC70. Quantitative real-time PCR analysis revealed that HSP70 mRNA was significantly upregulated by TDP-43 mNLS,C173S/C175S and by 3B12A scFv-CMA in the presence of either cytoplasmic TDP-43 or the WT TDP-43. The scFv-CL1 also increased HSP70 mRNA levels, but not HSP70 protein levels, only in the presence of TDP-43 aggregates. The induction of HSP70 mRNA was highest in cells cotransfected with 3B12A scFv-CMA and TDP-43 mNLS,C173S/C175S (Fig. 6d). Immunofluorescence microscopy also revealed that endogenous HSP70 colocalised with nuclear and cytoplasmic TDP-43 aggregates in HEK293A cells (Fig. 6e). Consistent with immunofluorescence results, an immunoprecipitation assay showed that 3B12A scFv-CMA augmented the interaction of HSP70 with misfolded TDP-43 in HEK293A cells (Fig. 6f). Notably, Western blot analysis revealed that HSP70 overexpression significantly reduced the detergent insolubility of cytoplasmic aggregated TDP-43, suggesting initiation of refolding, while overexpression of another representative molecular chaperone, HSP90, had no effect on TDP-43 solubility (Fig. 6g,h). As expected, expression of the constitutive protein HSC70 remained stable as evidenced by immunoblot analysis using an antibody recognising both HSC70 and HSP70 (Fig. 6a,c). An HSC70-specific antibody was not available. (c,d) Quantified data for HaloTag pulse chase assay of (a,b) respectively. Each data point was obtained by normalisation to actin. 3B12A scFvs promoted the degradation of the TDP-43 mNLS,C173S/C175S mutant. Moreover, VH_VL-CMA significantly enhanced degradation compared with VH_VL. Differences were evaluated by twoway ANOVA (mean ± SD from three independent experiments; **p < 0.01, ***p < 0.005 and ****p < 0.001). N.S. indicates not significant. (e,f) Quantified data for HaloTag pulse chase assay of TDP-43 mNLS and TDP-43 C173S/C175S , respectively. Differences were evaluated by two-way ANOVA (mean ± SD from three independent experiments; *p < 0.05 and ***p < 0.005). N.S. indicates not significant. (g) HaloTag pulse chase assay for cytoplasmic aggregated TDP-43 in the presence of lactacystin or bafilomycin. (h) Quantitative analysis of (g) Each data point was obtained by normalisation to actin. Differences were evaluated by one-way ANOVA (mean ± SD from three independent experiments; **p < 0.01 and ***p < 0.005 vs DMSO control). aggregates were distinguished from non-aggregates by diminishing the fluorescence intensity of diffusely distributed cytoplasmic GFP-TDP-43 signals and subsequently quantified with ImageJ software. Differences were evaluated by two-way ANOVA (mean ± SD from three independent visual fields; *p < 0.05, ***p < 0.005 and ****p < 0.001). (e,f) Effects of TDP-43 species and 3B12A scFv on N2a cell toxicity (e) and viability (f) DCS represents the C173S/C175S mutant. Differences were evaluated by one-way ANOVA (mean ± SD from triplicate; *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001). (g) R110/AFC ratio obtained from (e,f) Differences were evaluated by one-way ANOVA (mean ± SD from triplicate; ***p < 0.005 and ****p < 0.001). N.S. indicates not significant. Taken together, these findings indicate that 3B12A scFv-CMA facilitates the proteolytic clearance of misfolded TDP-43 and that clearance is further augmented by the aggregate refolding effect of induced HSP70.
To precisely evaluate the effects of 3B12A scFv-CMA on the clearance of misfolded TDP-43 in vivo, TDP-43 plasmids were electroporated into the cortex together with the 3B12A scFv-CMA or control plasmid plus mCherry expression vector to validate electroporation efficiency. As expected, GFP-TDP-43 WT distributed to nuclei in the CP and did not form detectable aggregates. Moreover, the relative fluorescence signal intensity, number and size of TDP-43 WT puncta in nuclei were not altered in the presence of 3B12A scFv-CMA compared with control cortices expressing GFP-TDP-43 WT alone (Fig. 7b-e). Alternatively, TDP-43 mNLS,C173S/C175S formed cytoplasmic puncta, which were markedly reduced in intensity, number and size by co-electroporation of 3B12A scFv-CMA (Fig. 7b,f-h).
To investigate whether 3B12A scFv-CMA expression affects early brain development, brains electroporated with the mCherry plasmid alone or plus the 3B12A scFv-CMA plasmid at E13.5 were subjected to immunohistochemistry at postnatal day 21 (P21). Similar to mice electroporated with the empty vector, those expressing 3B12A scFv-CMA were safely born at term, grew normally and exhibited age-appropriate behaviours until P21 (Supplementary Movie 2). Immunohistochemistry revealed that Myc-tagged 3B12A scFv-CMA distributed throughout the neocortex ( Supplementary Fig. S8), while neuron counts were similar to those of vector controls as revealed by NeuN immunostaining (Supplementary Fig. S9a). There was no abnormal proliferation of GFAP-positive astrocytes or Iba1-positive microglia indicative of tissue damage and reactive inflammation ( Supplementary Fig. S9b,c).

Discussion
We demonstrate that 3B12A scFv-CMA can serve as an effective and selective proteolytic intrabody against intracellular misfolded TDP-43 in cultured cells and murine cerebral cortex. A scFv targeting the N-terminal amino acids adjacent to the expanded polyglutamine of huntingtin protein was previously shown to successfully prevent protein aggregation 46,47 , underscoring the broad potential of this intrabody strategy for control of neurodegenerative diseases due to proteinopathy. Similarly, a scFv against amyloid oligomers attenuated α-synuclein aggregation and improved motor and cognitive functions in a mouse model of Parkinson's disease 48 . To the best of our knowledge, ours is the first study to evaluate the efficacy of a scFv intrabody with dual proteolytic signals for clearance of intracellular misfolded TDP-43 both in vitro and in vivo. Unexpectedly, the scavenging effect of CMA-fused intrabody was augmented by the aggregate refolding function of HSP70, which was transcriptionally induced upon aggregate−scFv binding (Fig. 6). The exact mechanism for HSP70 mRNA induction by the interaction between scFv-CMA and aggregated substrates remains elusive. Considering that the expression of TDP-43 aggregates or CMA-scFv alone did not evoke measurable HSP70 induction, autophagy induction might also be required. HSP70 confers neuroprotection by refolding TDP-43 49 , and the HSP70 inducer arimoclomol was tested in a clinical trial of ALS patients up to phase II/III. Therefore, the unique double action of our 3B12A scFv-CMA, proteolysis and refolding, is an effective mechanism for molecular targeting and clearance of pathogenic proteins. Further, this strategy was highly specific to the pathogenic form, as 3B12A scFv-CMA interacted with TDP-43 mNLS whether aggregated or nonaggregated but significantly accelerated degradation only of the aggregated form.
The 3B12A scFv intrabody contains a PEST-like peptide on the VH domain, which may also be involved in proteasomal degradation [30][31][32] . The other function of the PEST-like peptide is to promote folding by altering the overall charge of the scFv to negative 50 . Indeed, the calculated net charges of 3B12A-VH-VL and 3B12A-VH-VL analysis for detergent insolubility of TDP-43 in the presence of HSP70 or HSP90 overexpression in HEK293A. Cell lysates were separated into RIPA-soluble or -insoluble fractions after centrifugation for 20 min at 15,000 × g at 4 °C and subsequently eluted in 2% SDS sampling buffer for 5 min at 95 °C. DCS represents the C173S/C175S mutant. (h) Quantified densitometric analysis of (g) Each data point was obtained by normalisation to actin. The insolubility ratio is normalised insoluble TDP-43 to normalised soluble TDP-43. Differences were evaluated by one-way ANOVA (mean ± SD from three independent experiments; *p < 0.05 vs vector). Electroporation was performed at E13.5, and brains were fixed at E16.5 followed by immunohistochemistry. Scale bar = 100 μm. (c-h) Quantification analysis of (b) TDP-43 aggregates were distinguished from non-aggregates by setting the fluorescence cut-off for the quantification using ImageJ software. Each data point was obtained by normalisation to mCherry. Differences were evaluated by unpaired t-test (mean ± SD, n = 7 for TDP-43 WT  without the PEST-like sequence at pH 7 were only −3 and −1, respectively. The overexpression of CL1-or CMA-scFv with or without TDP-43 aggregates did not affect the autophagic or proteasomal activity. Although the sensitivities of proteasomal activity assays and autophagic flux assays are not high, results thus far suggest that binding to intrabodies with proteasome-targeting CL1 and lysosome-targeting CMA signals may be sufficient to clear TDP-43 aggregates without disrupting normal TDP-43 function or protein homeostasis. Previous studies have documented the efficacy of PEST-fusion intrabodies against pathogenic proteins such as huntingtin 46 and α-synuclein 51 . It was also shown that intrabodies can disrupt the subcellular distribution of target proteins, including by inhibition of nuclear−cytoplasmic translocation 52,53 . Although cytoplasmic nonaggregated TDP-43 (TDP-43 mNLS ) was not degraded by 3B12A scFv in cultured cells (Fig. 4e), further evaluations of 3B12A scFv effects on pre-aggregated TDP-43 nuclear−cytoplasmic redistribution are required to assess the potential for pre-emptive therapy.
The 3B12A antibody targets D247 of TDP-43 27 , a residue masked in the physiological condition by its location inside the molecule. The next residue, E246, is recognised by the von Hippel Lindau (VHL)/cullin-2 ubiquitin ligase complex for misfolded TDP-43 29 . This explains why 3B12A scFv specifically interacted with mislocalised/ misfolded TDP-43 but not with TDP-43 WT in cells (Fig. 2).
There is yet no model TDP-43 species replicating all pathological features of sporadic ALS. In this study, we used the C173S/C175S artificial mutant as a TDP-43 inclusion mimic based on evidence of structural similarity to aggregated TDP-43 inclusions in ALS 33 . Using high pressure NMR analysis, it was previously shown that C173 and C175 residues in the RRM1 domain are crucially involved in preserving the conformation of TDP-43, while disruption of the equilibrium between free and intramolecular disulfide states may cause irreversible TDP-43 aggregation. The structural similarity of this C173S/C175S mutant to TDP-43 inclusions was supported by immunohistochemistry using an antibody against the deformed loop (W113-T116) in RRM1 33 . The sustained mislocalisation of TDP-43 in the cytosol by the NLS mutation induces cell death and occasionally forms inclusions both in vitro and in vivo 34,54 . However, these aggregates are relatively small in number and rarely phosphorylated or ubiquitinated 34,54 ; thus TDP-43 mNLS is an insufficient model for TDP-43 proteinopathy. Transgenic mice overexpressing human TDP-43 WT and several known FALS mutants such as A315T and Q331K under native promoters develop similar mild phenotypes for FTLD-motor neuron diseases, but nuclear exclusion and cytosolic aggregates are prominent only in the FALS mutant models 55 . We reported that both WT and NLS mutant TDP-43 induce cell toxicity in immortalised cultured cells 34 . It should be noted that in the current work and our previous study 33 , aggregate-prone TDP-43 mNLS,C173S/C175S induced greater cell toxicity than TDP-43 WT or TDP-43 mNLS , irrespective of mislocalisation. The pathologies induced by this TDP-43 mutant share several crucial properties with TDP-43 proteinopathy in addition to the strong trend for aggregation, including the involvement of prion-like domains, impaired RNA splicing, ubiquitination and phosphorylation, as well as motor neuron toxicity 33 . Based on the extent of aggregate toxicity, the TDP-43 mNLS,C173S/C175S mutant may be the most robust form for modelling ALS-associated TDP-43 proteinopathy, especially for studies on therapeutic strategies and efficacy.
Most promising for clinical applicability, 3B12A scFv-CMA reduced cytoplasmic TDP-43 aggregates in the lateral CP of murine embryos (Fig. 7) without causing gross developmental defects. Indeed, mice expressing 3B12A scFv through in utero electroporation did not show aberrant/ectopic neuronal cells, gliosis or microglial activation compared with vector controls. Of course, the restricted distributions of exogenous human TDP-43 and 3B12A scFv-CMA protein in mice following in utero electroporation limit future experiments on mitigation of neurological behaviour phenotypes, and the environment of the pup brain is distinct from the aged human brain. However, the successful elimination of TDP-43 aggregates in vivo validates the basic principle and justifies further analyses in older disease model animals. In the clinical setting, our intrabody may be effectively delivered by a virus vector, such as an adeno-associated virus (AAV). Indeed, AAV9-mediated intrathecal delivery of a scFv derived from the D3H5 monoclonal antibody against misfolded SOD1 reduced misfolded SOD1 in the spinal cord, delayed disease onset and enhanced the survival of SOD1 G93A transgenic mice 56 . A recent study demonstrated that suppression of human TDP-43 mutant expression after disease onset in a mouse model decreased TDP-43 pathology, rescued motor deficits and prolonged survival 23 . These results provide strong support for our approach of pathological TDP-43 removal using a 3B12A intrabody, even after disease onset and formation of pathological TDP-43 aggregates. Further investigations evaluating the potential of our degradative 3B12A intrabody against pathogenic TDP-43 using appropriate animal models are required to advance this novel molecular targeting therapy for ALS.
Western blotting and immunoprecipitation. Cultured cells were lysed in sodium dodecyl sulfate (SDS) buffer containing 2-mercaptoethanol for Western blotting or in immunoprecipitation assay (RIPA) buffer (20 mM HEPES-KOH [pH 7.4], 125 mM NaCl, 2 mM EDTA, 1% Nonidet-P40, 1% sodium-deoxycholate) containing protease inhibitor cocktail (Roche, Basel, Switzerland) for immunoprecipitation. Ten percent of each cell lysate volume was analysed as the total cell lysate, and the remaining 90% was incubated with anti-FLAG M2 affinity gel (Sigma, St. Louis, MI) or anti-Myc affinity beads (Nacalai) at 4 °C overnight. The affinity beads were subsequently washed with RIPA buffer five times, and the immunoprecipitates were eluted in 2% SDS sampling buffer for 5 min at 95 °C. The eluates were separated on polyacrylamide gels (Wako, Tokyo, Japan) and proteins transferred onto PVDF membranes (Millipore, Billerica, MA). Proteins were detected using an enhanced chemiluminescence system (ECL; Thermo-Fisher Scientific, Waltham, MA or Nacalai). Densitometric analysis of protein bands was performed using ImageJ software 58 . Immunofluorescence and microscopic analysis. Cultured cells were fixed in 4% paraformaldehyde (PFA)/PBS (pH 7.2) and permeabilised with 0.1% Triton-X100/PBS containing 5% normal goat serum as a blocking agent. Cells were reacted with primary antibody (4 °C, overnight) and subsequently with a fluorophore-tagged secondary antibody (Alexa; Invitrogen) for 1 h at room temperature. Cells were counterstained with 4′-6 diamidino-2-phenylindole (DAPI). Fluorescence images were obtained using a confocal laser microscope (FV1000-D IX81, Olympus, Tokyo, Japan).
Sandwich ELISA. Target FLAG antibody in coating buffer (Roche) was coated onto ELISA plates (Nunc, Rochester, NY) at a 1:100 dilution. Coated FLAG antibody was then reacted with cell lysates from transfected HEK293A cells (prepared in RIPA buffer) for 1 h at room temperature and subsequently with anti-Myc antibody at a 1:500 dilution (4 °C, overnight). Then, the coated antibody was reacted with the peroxidase-conjugated secondary antibody (Jackson Immunoresearch, West Grove, PA). Finally, reaction buffer containing 2,2′-azino-bis-3-et hylbenzothiazoline-6-sulfonate (Roche) was applied, and the absorbance was measured by a spectrometer at 405 nm with a reference at 490 nm.
HaloTag pulse chase assay. At 48 h after cotransfection of HEK293A cells, the HaloTag-fused protein was labelled with 1 μM diAcFAM ligand (Promega) in FBS-free culture medium for 15 min in a cell culture incubator. Cells were then washed three times with warm PBS and cell lysates prepared in SDS sample buffer at 0, 12 and 24 h after ligand treatment. All lysate samples were separated on polyacrylamide gels (Wako) by SDS-PAGE, and gels were subsequently analysed on a fluorescence scanner (LAS-3000; FUJIFILM, Japan).

Time lapse imaging analysis. At 24 h after cotransfection, HEK293A cells were counterstained with
Hoechst 33342 (Nacalai). Stained cells were observed at 30-min intervals over 48 h using a fluorescent microscope (BZX-710; Keyence, Osaka, Japan). For quantification of GFP-positive TDP-43 aggregates with ImageJ software, we removed background noise and set fluorescence thresholds of GFP fluorescence at 2.5% to distinguish TDP-43 aggregates from diffusely expressing cytoplasmic non-aggregated species and subsequently counted GFP-positive areas over 10 µm 2 as aggregates. Intensity is expressed in arbitrary units.
Protein degradation chase assay. Protein half-life in HEK293A cells was estimated starting 48 h after transfection by chronological chase analysis. Cells were treated with CHX (100 μg/ml) to inhibit protein synthesis.

Measurement of cytotoxicity and cell viability.
We estimated the cytotoxicity and viability of transfected N2a cells using the MultiTox-Fluor multiplex cytotoxicity assay kit (Promega). Briefly, N2a cells were seeded onto 96-well black culture plates and plasmids encoding FLAG-tagged WT or mutant TDP-43 and 3B12A scFv were co-transfected at 0.1 µg each per well. At 48 h after transfection, cell-permeant glycyl-phenylalanylamino fluorocoumarin (GF-AFC) and cell-impermeant bisalanyl-alanyl-phenylalanyl-rhodamine 110 (bis-AAF-R110), fluorescent indicators for live cells and dead cells, respectively, were applied. After a 2-h incubation, relative fluorescence emission was measured on a multi-plate reader (Perkin Elmer) with excitation/emission of 400/505 nm for live cells and 485/520 nm for dead cells.
Brain slice preparation and immunohistochemistry. Embryos were anaesthetised on ice and postnatal mice by intraperitoneal injection of sodium pentobarbital and isoflurane inhalation. Brains were fixed by transcardial perfusion with 4% paraformaldehyde (PFA) in 0.1 M PBS. The brains were immersed in 4% PFA at 4 °C overnight, incubated in 20% sucrose/0.1 M PBS at 4 °C overnight, embedded in OCT compound (Sakura Finetek, Japan), and frozen in liquid nitrogen. The restricted distribution of the plasmids in the dorsolateral cortex of the pup mice by in utero transporation allowed us only limited slices in serial sectioning. We therefore carefully tried to obtain adjacent areas. Frozen sections were cut coronally and mounted on glass slides (Matsunami, Osaka, Japan). The slides were washed with 0.1 M PBS/0.1% Triton X-100 (PBS-T) and blocked with 3% bovine serum albumin (Nacalai) in PBS-T at room temperature for 30 min. After subsequent incubation with primary antibody at 4 °C overnight, slides were incubated in Alexa Fluor-conjugated antibody (Invitrogen) at room temperature for 1 h, washed with PBS-T, and placed onto cover glass (Matsunami) with mounting medium containing DAPI for nuclear counterstaining (Vector Laboratories, Burlingame, CA). For quantification of GFP-positive TDP-43 aggregates in the lateral cortical plates with ImageJ software, we removed background noise and set fluorescence thresholds of GFP fluorescence at 1% to distinguish TDP-43 aggregates from diffusely expressing cytoplasmic non-aggregates, and subsequently counted GFP-positive areas over 20 µm 2 as aggregates. Intensity is expressed in arbitrary units.

Statistical analysis.
Multiple group means were compared by one-way ANOVA with post hoc Tukey's multiple comparison tests for pair-wise comparisons. Factor estimation in the two chronological data groups was evaluated by two-way ANOVA using Prism software (GraphPad, La Jolla, CA). A p < 0.05 was considered statistically significant.