TDP-43 self-interaction is modulated by redox-active compounds Auranofin, Chelerythrine and Riluzole

Amyotrophic lateral sclerosis (ALS) represents a fatal neurodegenerative disease, which is characterized by a rapid loss of lower and upper motor neurons. As a major neuropathological hallmark, protein aggregates containing the Transactivating Response Region (TAR) DNA Binding Protein (TDP-43) are detectable in about 95% of sporadic ALS patients. TDP-43 interacts with itself physiologically to form liquid droplets, which may progress to pathological aggregates. In this study, we established the NanoBit luciferase complementation assay to measure TDP-43 self-interaction and found the fusion of the split luciferase subunits to the N-terminus of the protein as the strongest interacting partners. A screen of pharmacologically active compounds from the LOPAC®1280 library identified auranofin, chelerythrine and riluzole as dose-dependent inhibitors of TDP-43 self-interaction. Further analysis of drug action of the gold-containing thioredoxin reductase inhibitor auranofin revealed a redistribution from insoluble TDP-43 protein pool to PBS-soluble protein pool in N2a cells. In addition, auranofin treatment diminished reduced glutathione as a sign for oxidative modulation.

TDP-43 assemblies may represent a crucial intermediate state prior to TDP-43 aggregation 18 . Therefore, substances modulating TDP-43 self-interaction may interfere with formation of pathogenic oligomeric aggregates and hold therapeutic promise. For measurement of TDP-43 self-interaction, NanoBit luciferase complementation assay has been applied 19 . Two different NanoLuc luciferase subunits (large Bit (lgBit), an 18 kDa polypeptide, and small bit (smBit), a 1.3 kDa peptide) have been fused to the N-terminal and/or C-terminal site of TDP-43 protein, respectively. The NanoBit fusion-subunits only weakly associate by themselves (K D > 100 µM), so that their reconstitution of luciferase activity is controlled by the interaction characteristics of the tagged proteins 19,20 .
In our study, we found the highest self-interaction potential of constructs pFN33_TDP-43 and pFN35_TDP-43 with each of the NanoBit subunits fused to the N-terminal site of TDP-43 having a parallel TDP-43 orientation and allowing an undisturbed self-interaction of other regions like RRM domains or C-terminus of TDP-43. Testing pharmacologically active compounds from LOPAC ® 1280 library, we found riluzole, auranofin and chelerythrine as substances, which significantly decreased TDP-43 self-interaction. Especially auranofin showed impressive results with a straight, dose-dependent reduction of TDP-43 self-interaction without cell toxicity effects. Furthermore, auranofin seems to improve TDP-43 solubility as exemplified by a redistribution of TDP-43 from Triton X-100 and urea-soluble protein pool to PBS-soluble protein pool.

TDP-43 self-interaction.
For evaluation of the best configuration we tested four different pairings of TDP-43 fusion proteins. Fusion of both the large bit and the small bit to the N-Terminus of TDP-43 resulted in the highest luciferase activity (Fig. 1). Second best was the pairing where both luciferase bits were fused to the C-terminus to TDP-43, pointing to a parallel interaction of TDP-43. Antiparallel orientation of the luciferase bits yielded luciferase activity barely higher than the negative control ( Fig. 1, first column).
Screening of pharmacologically active compounds from LOPAC ® 1280 library on TDP-43 self-interaction (single point determination). LOPAC ® 1280 library, a collection of inhibitors, receptor ligands, pharma-developed compounds and approved drugs covering most signaling pathways and major drug target classes was applied at 10 µM concentration to the transfected mouse neuroblastoma Neuro2a (N2a) cells for the TDP-43 self-interaction assay. Screening the library compounds was performed as single determination in 384 wells. In order to reduce variability, we obtained a luciferase reading from each well after treatment and normalized to the luciferase reading of the same well before treatment (Fig. 2). Further, to detect off-target effects of compounds such as interference with luciferase reaction, protein homeostasis and TK promoter activity, we applied the same compound library onto N2a cells transfected with construct expressing full length constitutively-active NanoLuc luciferase driven by the same thymidine kinase promoter like the TDP-43 construct. Therefore, results of TDP-43 self-interaction were normalized with respect to NanoLuc activity after library treatment (Fig. 2). Compounds causing a reduction of TDP-43 interaction self-interaction to 70% [after%/pre%] or lower were considered as inhibitors and compounds, which increased the interaction to 120% [after%/pre%] or higher have been considered as stimulators (Table 1).
Dose-response relationship for selected compounds. In our validation study we observed a dose-dependent reduction of TDP-43 self-interaction for the substances riluzole, auranofin and chelerythrine (Fig. 3), while effects of the substances GABA, Oxotremorine sesquifumarate salt, Suramin sodium salt and Reactive Blue 2 have not been reproducible (Table 2). For the dose-response experiments we applied the compounds at final concentrations of 1.6 µM, 4.0 µM, 10.0 µM, 25.0 µM and 62.5 µM to both the TDP-43 interaction assay and the NanoLuc reference assay (Fig. 4). The WST-8 assay has been measured in the same wells for evaluation of cell viability immediately after luciferase readings (Supplemental Fig. 2). Riluzole treatment resulted in a significant reduction of TDP-43 interaction at concentrations of 4 µM (64% of untreated cells), 25 µM (56%) and 62.5 µM (42%), while Flowchart of NanoBit luciferase measurement with LOPAC ® 1280 -screen. Description of measurement procedure with compounds from LOPAC ® 1280 library. For reduction of variability, luciferase reading from each well after treatment has been normalized to the same well before treatment. To detect off-target effects of compounds such as interference with luciferase reaction, protein homeostasis and TK promoter activity, we applied the same compounds onto N2a cells transfected with a TK-promoter construct expressing full length NanoLuc luciferase. Therefore, results of TDP-43 self-interaction were normalized with respect to NanoLuc activity after library treatment.
Auranofin treatment at a concentration of 4 µM caused a non-significant reduction of TDP-43 self-interaction (45% of non-treated cells), whereas auranofin concentrations of 10 µM (15.4%), 25 µM (1.5%) and 62.5 µM (1.1%) had a strong and significant effect (Fig. 3C). After normalization of the TDP-43 interaction results to the constitutive NanoLuc, effects became more evident revealing that even the lowest concentration of 1.6 µM auranofin resulted in a significant reduction of 76% of non-treated cells (Fig. 3D).

Endogenous TDP-43 expression under treatment with riluzole, auranofin and chelerythrine.
Because the cellular amount of the LgBit-and SmBit-TDP-43 fusion protein is much lower than the endogenous TDP-43 content (Supplemental Fig. 3), we assume that endogenous TDP-43 present in N2a cells can affect the read-out of the NanoBit interaction assay by competing with the Plasmid-coded TDP-43 fusion proteins. In order to estimate a possible impact of the compounds on the steady-state level of endogenous TDP-43 protein, we performed an immunocytochemistry-based quantification. NanoLuc-only transfected N2a cells of the verification experiments have been fixed and permeabilized after luciferase and WST-8 measurement, and labeled with anti-TDP-43 antibody and Hoechst 33342 (Fig. 5). This procedure allowed us to estimate changes in nuclear TDP-43 by fluorescence densitometry after pharmacological treatment in Hoechst 33342 positive nuclei ( Fig. 5A,C,E). We observed an increase in endogenous TDP-43 at low compound concentrations such as 1.6 µM riluzole (37% increase, non-significant), 1.6 µM auranofin (88% increase, p < 0.01) and 4 µM chelerythrine (38%, p < 0.01, Fig Fig. 5E) -probably due to its cell toxicity, but did not show a dose-dependent effect on relative TDP-43 expression (Fig. 5F). Taken together, these results lead us to conclude that slightly changes in the TDP-43 steady state levels are unlikely to cause the reduction of TDP-43 self-interaction of treated N2a cells.

Discussion
TDP-43 is a major component of cytoplasmic proteinaceous aggregates and detectable in about 95% of sporadic ALS patients 6,7 . Thus, TDP-43 aggregation is seen as a neuropathological hallmark of ALS 4 . Beside ALS, TDP-43 inclusions have been observed in a broad spectrum of neurodegenerative disorders such as ALS/ parkinonism-dementia complex of Guam, Alzheimer disease, dementia with Lewy bodies, Pick's disease, argyrophilic grain disease and corticobasal degeneration 24 . The TDP-43 aggregation appears to be cooperatively mediated by several protein regions like N-terminus, RNA recognition motif (RRM) domains and C-terminus 11,16,17 .
In our study, we analyzed the TDP-43 protein-protein self-interaction using NanoBit complementation reporter assay, which allows sensitive detection of protein-protein interaction with low expression levels of exogenous split fusion proteins based on luminescence compared to bimolecular fluorescence complementation assay (BiFC) based on fluorescence 19,25 . The highest self-interaction potential has been observed in the constructs pFN33_TDP-43 and pFN35_TDP-43 with each of the NanoBit subunits fused to the N-terminal site of the We tested modulation of TDP-43 self-interaction by pharmacologically active compounds from LOPAC ® 1280 library and found a reproducible, dose-dependent reduction of TDP-43 self-interaction by auranofin without reduction of endogenous TDP-43 expression and without cell-toxicity effects.
Beside some false hits from our initial screen we detected few substances which diminished TDP-43 self-interaction in our validation study such as Riluzol, auranofin, chelerythrine and sanguinarine (not shown). In our further analysis, we concentrated on auranofin, an organic gold thiol compound, because of its strong inhibition and good toxicity characteristics and due to its clinical use in therapy of rheumatoid arthritis to reduce inflammation 23,26 . It acts as a thioredoxin reductase inhibitor and competitively binds reduced selenocysteine located within the enzyme redox center 23 . Specifically, it is thought that the C-terminal redox-active Cys-495/ SeCys-496 center of thioredoxin reductase is the target of auranofin 23,27 . Six cysteine residues have been also described in TDP-43 protein, four of which (Cys173, Cys175, Cys198 and Cys244) are located in the two RRM domains, while the other two (Cys39 and Cys50) are in the N-terminal domain 28 . Oxidation of cysteines located in the two RRMs decreases protein solubility, and leads to the formation of intra-and inter-molecular disulfide linkage 29,30 . Furthermore, cysteine residues in RRM1 may direct the conformation of TDP-43 31 . Thus, auranofin may be involved in regulation of TDP-43 self-interaction by modulating the cysteine residues in RRM domains of TDP-43. Beside this, auranofin induces a striking intracellular redistribution of TDP-43 from particulate protein fraction to the PBS-soluble protein fraction as shown by sequential protein extraction. This was obvious in Triton X-100 extractable particulate protein fraction, which may represent luminal and membrane-bound protein from cell membrane and organelles such as mitochondria and endosomes or even membrane-less liquid-like droples 32 , in which TDP-43 is enriched. In the subsequent 8 M urea extractable protein pool again TDP-43 was largely depleted after auranofin treatment. This protein pool comprises nuclear matrix protein, cytoskeletal proteins and protein complexes that resisted preceeding extractions. Thus, auranofin seems to release its inhibitory effect on TDP-43 self-interaction either by shifting it from membrane-bound organelles to the cytoplasm, by dissolving TDP-43 containing liquid-like droplets or by diminishing TDP-43 from subnuclear compartments. This finding is complemented by the size-exclusion chromatography results for PBS-soluble TDP-43, which shows that the molecular weight of the main TDP-43 peaks remain unchanged but that the 67 kDa peak is increased.
Interestingly, auranofin has been shown to reach CNS concentrations above 0.1 μM after oral administration in rats and mice 26 suggesting that pharmacokinetic properties of auranofin may allow to explore its therapeutic efficacy in ALS. Further, because auranofin has redox-modulating properties, it would target an additional pathogenetically important mechanism as indicated by the disease-modifying effect of free-radical scavenger edaravone in some ALS patients 33 .
Beside auranofin, riluzole showed a significant inhibition of TDP-43 self-interaction without cell toxicity and without change in TDP-43 expression. At present, riluzole is the only drug in Europe approved for treatment of ALS based on its moderate increase of survival time by 3 months [34][35][36] . Mechanistically, riluzole is thought to have a broad spectrum of action with reduction of excitotoxicity by the excitatory amino acid neurotransmitter glutamate by enhancing glutamate uptake through glutamate transporters GLAST, GLT1 and EAAC1 35,37 , inhibition of voltage-dependent sodium channels 38 and potassium channels 39 and inhibition of protein kinase C, suggesting its involvement in antioxidative processes 40 . Further, oxidative modulation of cysteine residues during regulation auf K + channels has been reported 41 . The present findings appear to be the first to demonstrate modulation of TDP-43 self-interaction suggesting that inhibition of TDP-43 self-interaction may be an additional mechanism by which riluzole may exert its beneficial effect in ALS. Because riluzole has antioxidative properties, it may inhibit TDP-43 aggregation by a mechanism similar to auranofin.
Finally, we found chelerythrine to be a strong inhibitor of TDP-43 self-interaction. While its application did not result in a reduced TDP-43 expression, substantial cell toxicity was found in concentrations of 10 µM and above. Chelerythrine is a benzophenanthridine alkaloid from the plant Chelidonium majus. It acts as a potent, selective and cell-permeable protein kinase C inhibitor 42 and as antagonist of G-protein-coupled cannabinoid CB 1 receptors 43 . Further, Chelerythrine induces production of reactive oxygen species (ROS) 44,45 and participates in reversible complexation with cysteine 46 . Thus, chelerythrine may be involved in oxidative modulation of cysteines located in the RRM regions of TDP-43 as well. In addition, a close chelerythrine analogue sanguinarine with similar pharmacological properties has been identified as a potential TDP-43 interaction modulator in our initial screen (see Table 1).
In conclusion, we established the NanoBit complementation reporter assay to measure TDP-43 self-interaction and found two TDP-43 constructs with N-terminal tag as the strongest interacting partners. A screen of pharmacologically active compounds from the LOPAC ® 1280 library identified auranofin, riluzole and chelerythrine as inhibitors of TDP-43 self-interaction without reduction of TDP-43 expression. Pharmacological properties of all three substances and preliminary findings for auranofin may point to the hypothesis that inhibition of TDP-43 self-interaction is due to oxidative modulation of cysteine residues in RRM domains of TDP-43.

TDP-43-CTF 5′-GCAGAGGGAGCCAAACCAGG-3′
Nano-R 5′-GTTAGCAGCCGGATCAGCTTG-3′ SgfI/PmeI and C-terminal fusion vectors (pFC34, pFC36) digested with SgfI/EcoICRI, so TDP-43 is fused in frame with large bit of the split luciferase (pFN33, pFC34) or the small bit (pFN35, pFC36). Ligated constructs were transformed into high-efficiency E. coli competent cells (C2987H) and plated on kanamycin LB plates. The plasmids were purified from single colonies by QIAprep ® Spin MiniPrep Kit (Qiagen, Hilden, Germany) and NucleoBond ® Xtra Midi EF Kit (Macherey-Nagel, Düren, Germany). Correct coding sequence has been controlled by sequencing with sequencing primers (Table 3).   Sequential protein extraction of TDP-43. N2a cells have been cultured in 6 cm dishes in triplicates either untreated, treated with 0.1% DMSO as a vehicle control or 10 µM auranofin for 60 min at 37 °C. Before scraping cells off the dish and pelleting cells at 400 × g and 4 °C for 10 min cells were washed with cold PBS. Each pellet contained about 17 mg of cell mass for each condition. Soluble cytosolic proteins were extracted with 200 µl PBS + protease and phosphatase inhibitors (cOmpleteTM Protease Inhibitor Cocktail, PhosSTOP, Roche, Basel, Switzerland) assisted by ultrasonication using three 2-second bursts at 40% maximum output (Virsonic cell disrupter) followed by ultracentifugation at 100,000 × g at 4 °C for 30 min in Sorvall MTX150 using S140AT rotor (ThermoFisher Scientific, Schwerte, Germany). TDP-43 contained in the insoluble particulate protein fractions was sequentially solubilized by 150 µl 1% Triton X-100 in PBS buffer (Triton X-100 soluble fraction) during a 60 min incubation at 4 °C followed by ultracentrifugation at 100,000 × g at 4 °C for 30 min. Thereafter, pellets were incubated with 80 µl 8 M urea containing PBS buffer (urea-soluble fraction) overnight at room temperature followed by ultracentrifugation. The resulting final pellets were resuspended in 20 µl 70% formic acid and agitated for 2 h at room temperature.