Levosimendan prevents tau pathology by inhibiting disulfide-linked tau oligomerization posing as a promising anti-tau therapeutics


 Tau oligomers play critical roles in tau pathology, responsible for neuronal cell death and transmitting the disease in the brain. Accordingly, preventing tau oligomerization becomes an important therapeutic strategy to treat tauopathies including Alzheimer’s disease, however progress has been slow due to difficulties of detecting tau oligomers in cellular context. Toward tau-targeted drug discovery, our group have developed a tau-BiFC platform to monitor and quantify tau oligomerization. By using the tau-BiFC platform, we screened 1,018 compounds in FDA-approved & Passed Phase I drug library, and identified levosimendan as a potent anti-tau agent inhibiting tau oligomerization. 14C-isotope labeling of levosimendan identified that levosimendan covalently bound to tau cysteines, directly inhibiting disulfide-linked tau oligomerization. In addition, levosimendan was able to disassemble tau oligomers into monomers, and rescuing neurons from aggregation states. In comparison, the well-known anti-tau agents, methylene blue (MB) and LMTM, failed to protect neurons from tau-mediated toxicity, generating high-molecular weight tau oligomers. The administration of levosimendan also suppressed tau pathology in the brain, preventing cognitive declines in TauP301L-BiFC transgenic mice. Although careful validation is required, here we present the potential of levosimendan as a disease modifying therapy for tauopathies targeting tau oligomerization.


Introduction
Naturally, tau is an extremely soluble protein containing a number of positively charged lysine residues 1 .
The lysine residues play a role for binding negatively charged microtubules 2 . Upon binding, tau stabilizes microtubule and promotes microtubule assembly, which is critical for axonal outgrowth 3,4 . Under pathological condition, the charge balance between tau and microtubule is disrupted, and tau dissociates from microtubules 5 . The dissociated tau become a susceptible substrate of cytosolic enzymes that modify the states of tau phosphorylation, acetylation, or proteolytic cleavage 6,7 . Chemically and structurally modi ed tau becomes aggregated and accumulated in neurons. Intraneuronal deposits of tau brils (PHFs and NFTs) are the pathological hallmark in a group of neurodegenerative diseases, called tauopathies. Tauopathies include not only Alzheimer's disease (AD), but also Fronto-Temporal Dementia (FTD), progressive supranuclear palsy, corticobasal degeneration, and chronic traumatic encephalopathy 8,9 . Due to the implication of tau in neurodegenerative disorders, tau has become an important therapeutic target 10 . However, progress has been slow due to the lack of understanding tau pathology.
Until recently, neurotoxic tau species were presumed to be the lamentous tau (PHFs and NFTs), accumulated in the brain of AD patients. However, mounting evidence suggests that soluble tau oligomer, rather than brils, is neurotoxic species responsible for neuronal cell death 11,12 . Tau oligomers are soluble forms of tau aggregates ranging from dimers to pre brillar aggregates 13 . In AD brains, tau oligomers are detected in the early stage of pathogenic cascades, and the level of tau oligomers correlated with synaptic dysfunction and neuronal cell loss, rather than the levels of NFTs [14][15][16] . Evidence also suggested that tau oligomers are transmittable between neurons, spreading the disease in the brain 17,18 . Due to its relatively small size, tau oligomers are able to penetrate into cells and initiate nascent tau aggregation, providing structural seeds for aggregation 19 . The formation of insoluble lament (PHFs and NFTs) might be a survival mechanism for neurons to quarantine toxic tau species from cytosol 20 . Accordingly, prevention of tau oligomerization or elimination of oligomers became an important therapeutic strategy for AD drug development.
The most studied tau aggregation inhibitor is methylene blue (MB), a blue dye, which has been widely used in various industrial and research settings. In 2008, Wischik et al. rstly reported the anti-tau activity of MB 21 . In their study, MB disrupted the stability of NFTs isolated from AD patients. In 2013, Zweckstetter et al. reported the inhibitory mechanism of MB on tau aggregation. In their study, MB inhibits tau aggregation through the oxidation of tau cysteine residues, converting tau into aggregationincompetent monomer 22 . However, there are also con icting data showing that MB increased tau dimers through the same cysteine-oxidation mechanism 23 . Full length tau contains two cysteine residues (C291 and C322) in the microtubule-binding domain. Intra-molecular disul de bond formation generates a compact monomer, resistant to tau aggregation, and inter-molecular disul de bonds facilitate the aggregation cascade by generating structurally stable tau oligomers 24 . In 2019, Takashima et al. reported that MB inhibited the formation of tau brils, but not tau oligomers 25 . While the mechanism of action on tau pathology is still unclear, the reduced form of MB, LMTM (also known as TauRx0237 or LMT-X), has completed Phase III clinical trials for the treatment of Alzheimer's disease. In rst Phase III trial, LMTM failed to ameliorate cognitive decline in patients with mild to moderate Alzheimer's disease 26,27 . Although failed in AD treatment, LMTM is still on clinical trials for the treatment of FTDs.
Tau aggregation is a complicated multi-step process, which is controlled by a number of cellular enzymes. Therefore, in many cases, tau aggregation inhibitors which show strong inhibitory effects in in vitro tau aggregation assays, are not effective in cellular system. Therefore, the action-mechanism of tautargeted drugs should be evaluated in cellular system, not in test tubes using puri ed tau. In intracellular space, diverse tau modi cations occur simultaneously, changing its physiological properties; tau phosphorylation reducing its microtubule binding a nity 28, 29 , tau acetylation preventing ubiquitinmediated degradation 30 , and dityrosine or disul de-bond formation promoting tau aggregation by generating covalently linked tau oligomers 24,31 . In this regard, our group has developed tau-BiFC platform to monitor tau oligomerization in living cells and in the brain of mice 32,33 . In tau-BiFC system, nonuorescent N-and C-terminal compartments of Venus protein are fused to tau and Venus uorescence turns on only when tau assembles together (Fig. 1a). By using tau-BiFC cell model, we have characterized prion-like tau oligomers 19 , and investigated pathological tau modi cation [34][35][36] . Tau-BiFC cell screening identi ed that levosimendan is an effective drug in preventing FK-induced tau aggregation and also FK-induced cellular toxicity. However, FK activates not only tau kinases, but also other cellular processes activated with the increased level of cAMP 38 . The inhibitory effect of levosimendan could be the result of regulating the upstream cAMP pathway, rather than tau pathology.
To scrutinize the drug effect on tau pathology, tauK18 P301L was used to activate tau aggregation. TauK18 fragment containing P301L mutation, tauK18 P301L , activates tau aggregation as a prion-like seed 19 .
Again, levosimendan prevented tauK18 P301L -induced tau aggregation and cellular toxicity almost completely (EC 50 =2.2±0.1 mM, GI 50 =101.3±0.4 mM) ( Fig. 1g and Supplementary Fig. 2 (Fig. 1j). Upon the treatment of tauK18 P301L , intracellular tau aggregation increased signi cantly showing thread-like phenotype, and the increased BiFC uorescence was suppressed almost completely by the treatment of levosimendan. As shown in ow cytometry analysis, MB and LMTM was not effective in reducing intracellular tau aggregation.
Next, immunoblot analysis was followed to evaluate tau phosphorylation and total tau levels in tau-BiFC cells. To investigate dose-dependency, tau-BiFC cells were treated with increasing concentrations of each drug in the presence of tauK18 P301L (MB and LMTM; 0.5, 1.5, 5 mM, and levosimendan; 5, 15, 45 mM).
Upon the treatment of tauK18 P301L , the amount of total tau protein was increased to 2.0±0.1-fold compared to that of basal (Fig. 2c). TauK18 P301L -induced tau accumulation corresponds to the previous report showing that tau is metabolized at much slower rates when pathologically activated 39 .
Levosimendan treatment effectively suppressed tauK18 P301 -induced tau accumulation in tau-BiFC cells, keeping total tau level at 1.0±0.2-fold, which is comparable to the level of basal. Next, we investigated the formation of disul de-dependent tau oligomers on non-reducing SDS-PAGE gels. Substantial amount of disul de-linked tau oligomers were detected even at basal condition and the level of disul de-linked tau oligomers increased to 2.4±0.4-fold by the treatment of tauK18 P301L (Fig. 2d). Levosimendan treatment decreased the amount of disul de-linked tau oligomers, comparable to the basal level (0.8±0.2-fold at 45 mM). In contrast, MB and LMTM increased the level of disul de-linked tau oligomerization intensely (MB; 1.6±0.3-fold and LMTM; 1.9±0.3-fold at 5 mM), presenting highly enriched bands above 245 kDa (Fig.  2d). Non-reducing gel analysis indicates that levosimendan is effective in inhibiting tau-BiFC responses through the suppression of disul de-linked tau oligomerization. In comparison, MB and LMTM did not inhibit tau-BiFC responses and promoted the formation of structurally stable tau oligomers that are SDSand reducing agent resistant. However, careful validation was necessary since it was possible that MB and LMTM might act on BiFC complementation process instead of tau oligomerization.
Next, we evaluated the drug effect on tau oligomerization in primary neuron culture. Primary hippocampal neurons were isolated from day18 rat embryos. After 10 days of in vitro culture, the hippocampal neurons were treated with MB, LMTM (3 mM) or levosimendan (10 mM) in the presence of tauK18 P301L . In case of MB and LMTM, severe neuronal cell death was observed at higher concentrations over 10 mM, similar to those toxicity on tau-BiFC cells (data not shown). The neuronal cell lysates were subjected for immunoblot analysis in both reducing-and non-reducing-condition. Upon the treatment of tauK18 P301L , tau phosphorylation increased 2.0±0.2-fold at S396 and 1.7±0.1-fold at S199, indicating activated tau pathology in primary neuron (Fig. 2e, f). Again, levosimendan treatment suppressed tauK18 P301L -induced tau phosphorylation comparable to the basal level, showing 0.9±0.2-fold at S396 and 0.9±0.1-fold at S199. Also, in the primary neurons, the treatment of MB and LMTM increased the level of high-molecular weight tau oligomers resistant to SDS-and b-mercaptoethanol, while decreasing the level of monomers ( Fig. 2e, f). Upon the treatment of tauK18 P301L , the level of total tau increased to 1.5±0.1-fold compared to that of basal, indicating the accumulation of pathological tau species in neurons (Fig. 2g). Levosimendan treatment also suppressed tau accumulation in neurons, maintaining total tau level at1.0±0.1-fold comparable to that of basal.
Non-reducing SDS-PAGE analysis also indicated that, substantial amount of tau exists as disul de-linked tau oligomers in primary neurons (Fig. 2h). In basal condition, the level of disul de-linked oligomers was 2.4±0.7-fold, compared to the level of monomer. Upon tauK18 P301L activation, the amount of disul delinked tau oligomers was increased by 1.8±0.3-fold compared to that of basal. Levosimendan e ciently suppressed the formation of disul de-linked oligomers, decreasing the level of oligomers to 0.5±0.1-fold, which is even lower than that of basal. In case of MB and LMTM, the level of oligomers was increased to 2.3±0.5-and 2.4±0.3-fold, respectively, and monomers were almost disappeared (Fig. 2h). Our results clearly indicated that while MB and LMTM increases the formation of tau oligomers in primary neurons, levosimendan prevented tau phosphorylation, inhibiting disul de-linked tau oligomerization.
Non-reducing SDS-PAGE analysis indicated that signi cant amount of tau exists as disul de-linked oligomers upon the activation of tau pathology. Full-length human tau has two cysteine residues (C291 and C322) that can form intra-and inter-molecular disul de bonds 40 . While intramolecular disul de bonds lead to the formation of compact monomers that cannot form extended structure, intermolecular disul de-bonds produce covalently linked oligomers. The disul de-linked oligomers serve as ''nuclei'' for further tau aggregation 24 (Fig. 2i). Tau aggregation may occur in the absence of disul de-bond formation, but disul de-linked tau oligomers could facilitate tau aggregation, serving as a structural seed for tau aggregation. Our result indicates that the treatment of MB and LMTM increased the level of disul de-linked tau oligomers and levosimendan prevented the formation of disul de-linked tau oligomers.
Next, we evaluated the effects of tau oligomers on neuronal integrity. At DIV10, hippocampal neurons were treated with MB, LMTM (3 mM) or levosimendan (10 mM) followed by tauK18 P301L . After 48 hrs of incubation, neurons were stained with NeuO, a neuron speci c uorescence dye 41 (Fig. 2j). Then, uorescence images were acquired and analyzed to evaluate the length of neurites (Fig. 2k).
TauK18 P301L -induced neuronal degeneration was observed signi cantly by showing 142.8±17.8 mm shortened neurites compared to that of basal (p < 0.0001). Levosimendan inhibited tauK18 P301L -induced neuronal degeneration by recovering neurite length to 346.8±20.7 mm. In contrast, MB did not show signi cant effect on neuronal integrity and LMTM protected neuronal degeneration slightly by showing 151.1±24.4 mm neurite length. This result clearly indicates that levosimendan prevented tauK18 P301Linduced toxicity not only in tau-BiFC cells, but also in primary neurons, via inhibiting the formation of disul de-dependent tau oligomers.
Next, we validated the direct effect of each drug on tau oligomerization and tau brilization in in vitro conditions using puri ed tauK18 fragment. For tau oligomerization analysis, tau pre-aggregates were incubated with each drug at various concentrations (10, 30, and 100 mM) for 5 hours. Then, tau oligomers in the mixtures were separated into reducing and non-reducing SDS-PAGE ( Fig. 3a, b). In DMSO-treated control, 50±1.1% of total tau exists as disul de-linked oligomers, which were formed spontaneously in phosphate-buffered saline (pH 7.4). Upon the treatment of levosimendan, the levels of disul de-linked tau oligomers was decreased to 15.0±6.0% at 10 mM, 15.7±7.8% at 30 mM, and 12.3±15.0% at 100 mM, increasing the level of monomers. This result indicates that levosimendan directly inhibits disul de-linked tau oligomerization. In case of MB and LMTM, at 100 mM concentration, the level of disul de-linked tau oligomers was increased to 77.6±2.0% and 79.2±0.8%, respectively. MBand LMTM-induced tau oligomers are not easily dissociable into monomers even in reducing condition, showing noticeable bands of dimers and trimers, resistant to SDS-& b-mercaptoethanol on a reducing SDS-PAGE gel (Fig. 3b). In vitro tau oligomerization assay indicates that levosimendan prevents disul delinked tau oligomerization by direct interaction with tau protein, while MB and LMTM increase disul delinked tau oligomerization.
Then, we investigated the inhibitory effect on tau aggregation. For tau aggregation assay, tau aggregation was induced by the treatment of heparin to puri ed tauK18 in the presence of each drug at various concentrations for 5 days. The formation of b-sheet aggregates was evaluated with Thio avin S (ThS) assay. All drugs inhibited the formation of ThS-positive aggregates at micromolar concentration (MB; EC 50 =3.1±0.1 mM, LMTM; EC 50 =1.9±0.2 mM, and levosimendan; EC 50 =2.6±0.1 mM) (Fig. 3c). Then, the structures of tau aggregates were evaluated under transmission electron microscopy (TEM) (Fig. 3d).
TEM image of DMSO control shows the formation of long and straight laments of tau. In comparison, tau laments are not observed in levosimendan-treated mixture, indicating its effectiveness in inhibiting tau aggregation. In case of MB and LMTM, a few thread-like laments were observed. We also investigated the activity of drugs on reversing tau aggregation. For disaggregation assay, tau aggregates were treated with each drug at various concentrations for 4 days.  (Fig. 3e). However, ThS results of MB and LMTM con icted with TEM images showing thick bundles of tau laments in MB-and LMTM-treated mixture (Fig. 3f). The result corresponds with a previous study showing, upon treated to tau laments, MB did not reduced the tau laments shown under electron microscope 42 . Our result implies that the treatment of MB or LMTM prevents the interaction between ThS and tau laments, generating ThS-negative tau laments. In case of levosimendan, ThS response correlated with high-resolution TEM image, which shows a bunch of short tau fragments. The lengths of tau laments are ranging from 8.8 nm to 139.3 nm (n=403). The in vitro tau aggregation and disaggregation assay show the effectiveness of levosimendan as an anti-tau oligomerization agent, which not only inhibits tau aggregation, but also disassembles tau laments.
To further con rm the effectiveness of levosimendan on tau disassembly, levosimendan was treated at diverse time-points of tau aggregation processes in vitro and in tau-BiFC cells. For in vitro assay, tauK18 aggregation was induced by heparin, and the level of tau aggregation was monitored with ThS over 5 days. Levosimendan was treated to an aliquot of the aggregation mixture at 0, 25, 44, and 75 hours after heparin activation (Fig. 3g). ThS-response curve shows that tau aggregation followed nucleationelongation mechanism. Levosimendan treatment, no matter when it was treated, decreased ThS-positive tau aggregates (Fig. 3h). For tau-BiFC assay, tau pathology was activated by the treatment of tauK18 P301L , and tau-BiFC uorescence intensities were monitored over three days. Tau-BiFC cells were treated with levosimendan at 1, 12, 24, and 36 hours after tauK18 P301L activation (Fig. 3i). Again, levosimendan treatment, no matter when it was treated, decreased tau-BiFC uorescence responses, indicating the effectiveness as an anti-tau oligomerization agent ( Fig. 3j and Supplementary Fig. 3).

Levosimendan inhibits disul de-crosslinked tau oligomerization in vitro
Next we investigated the molecular mechanism of levosimendan in inhibiting oligomerization. We hypothesized that levosimendan inhibits tau-disul de bond formation by blocking tau cysteine residues, since its nitrile group is able to form a thioimidate bond with cysteine. To verify its covalent modi cation to tau, one of two nitrile groups of levosimendan was labelled with 14 C-radioisotope ( Fig. 4a and Supplementary materials). 14 C-levosimendan was treated to tau pre-aggregates that contains disul delinked tau oligomers for 2 hrs. Then, disul de-linked tau oligomers were separated on a non-reducing SDS-PAGE gel (Fig. 4b, c). 14 C-radiography showed tau monomers and dimers labelled with 14 Clevosimendan, indicating its covalent attachment to tau (Fig. 4c). Then, the relative intensities of Coomassie blue-stained bands of monomer, dimer, trimer, and tetramer were compared (Fig. 4c, d).
Corresponding to Figure 3e, the treatment of 14 C-levosimendan disassembled disul de-linked tau oligomers, increasing the level of monomers as levosimendan did (Fig. 4c, d). In addition, high-resolution TEM images were acquired to validate the size of tau oligomers in DMSO-treated control and levosimendan-treated mixture. Heterogeneous tau particles with diameters of 13.7±5.2 nm were observed in DMSO-treated control (n = 60) (Fig. 4e, f). The diameters of tauK18 oligomers matches with a previous report showing that tauK18 formed spherical tau oligomers in size with diameters of 10~20 nm 43 . In comparison, comparably small tau particles with diameters of 8.8±2.0 nm were observed in levosimendan treated mixture (n = 69), supporting the disassembly of tau oligomers by levosimendan. To validate levosimendan-binding to tau cysteine residues, each repeat domain of tauK18 was synthesized and incubated with levosimendan (Fig. 4g). MALDI-mass spectrometry indicated that levosimendan covalently attached to R2 and R3 domains containing a cysteine residue.
Next, we con rmed the importance of the nitrile moiety of levosimendan in inhibiting tau aggregation by using its metabolites. In human, 5% of levosimendan is metabolized to OR-1855 and OR-1896 in which the nitrile moiety is absent 44 (Fig. 4h). In particular, OR-1896 is a pharmacologically active metabolite of levosimendan, acting as a powerful inodilator 45 . The metabolite's effects on tau aggregation were evaluated in vitro and in tau-BiFC cells. For in vitro assay, tau aggregation was induced by heparin in the presence of each metabolite for 5 days. ThS responses indicated that OR-1855 and OR-1896 did not inhibit tau aggregation in vitro, while levosimendan showed strong anti-tau aggregation activity (Fig. 4i).
For tau-BiFC assay, tau-BiFC cells were treated with each metabolite for 48 hrs in the presence of tauK18 P301L . Again, even at 100 mM concentration, OR-1855 and OR-1896 did not inhibit the tau-BiFC uorescence increase (Fig. 4j). This result indicated that the nitrile group is critical in inhibiting tau aggregation through the modi cation of tau cysteine resides. Interestingly, OR-1896, which is a pharmacologically active metabolite acting as an inodilator, did not show any effect on tau. This result strongly suggests that the anti-tau activity is a new mode of action of levosimendan, distinguished from its known action as an inodilator.
Levosimendan rescues cognitive de cit and tau pathology in Tau P301L -BiFC mice.
Next, we evaluated in vivo e cacy of levosimendan in Tau P301L -BiFC mice 33 . Tau P301L -BiFC mouse model expresses human mutant (P301L) tau labeled by BiFC compartments (Fig. 5a). In the brain of Tau P301L -BiFC mouse, tau oligomerization occurs from 3 months and neuronal degeneration occurs from 9 months, leading to cognitive de cits at 12 months of age. To evaluate the therapeutic effect of levosimendan in preventing neuronal degeneration and cognitive impairment, Tau P301L -BiFC mice were received intraperitoneal administration of levosimendan or LMTM (5 mg/kg, three times per week) from 9 months to 12 months (Fig. 5a). After 4 month of drug administration, mice were subjected to behavioral tests to monitor cognitive function; novel objective test (NOR), Y-maze test, and passive avoidance test. For comparison, age-matched wild-type (WT) littermates were also subjected to the tests. In the novel objective recognition test, cognitive abilities were determined by the recognition index (RI) for a novel object. Vehicle-treated transgenic (TG) mice showed a signi cant decrease in the recognition index, with a RI value of 0.49 (p < 0.0001, compare to WT), while WT mice exhibited 0.74 of a RI value for a novel object. Levosimendan-treatment signi cantly improved recognition memory performance with a RI value of 0.69 (p < 0.0001). LMTM-treatment also attenuated the recognition de cit of TG mice by showing a RI value of 0.59 (p < 0.01) (Fig. 5b). In Y-maze test, cognitive ability was determined by the percent alternation. Each levosimendan-and LMTM-treated group showed signi cant increase in the alternation (70.4±4.2%, p < 0.05 and 68.9±3.5%, p < 0.05 compared to vehicle-treated, respectively), compared with the vehicle-treated group (52.8±2.6%, p < 0.01 compared to WT) (Fig. 5c). Further, passive avoidance test was followed to assess emotion-associated learning ability. Learning abilities were determined with the latency of entering the dark compartment, where the mice received an electrical shock a day before. Most mice in the vehicle-treated group entered the dark chamber without hesitation (p < 0.05), indicating the impairment of fear memory. Levosimendan-treated mice exhibited improved memory showing delayed latency or not entering to the chamber compared with the vehicle-treated TG controls (p < 0.05) (Fig. 5d). Interestingly, the LMTM-treated group, did not show signi cant improvement similar to vehicle-treated TG mice. All these results indicate that administration of levosimendan ameliorates tauopathy-induced memory de cits of aged, symptomatic tau TG mice.
Next, brains were extracted from mice, and brain tissue sections were prepared to evaluate tau pathology. Tau-BiFC uorescence images indicate the level of tau assembly including tau oligomers and aggregates and AT8-immunostain indicates the level of tau phosphorylation associated with late-state of tau aggregation. Tau-BiFC intensities decreased both in LMTM and levosimendan-treated groups (Fig. 5e, g). In levosimendan treated group, tau-BiFC intensities decreased to 0.6±01-fold in cortex and 0.4±0.2-fold in hippocampus, compared to those of vehicle-treated group. In LMTM-treated group, tau-BiFC intensities decreased to 0.7±0.2-fold in cortex and 0.6±0.1-fold in hippocampus. AT8-immunoreactivity also decreased in both LMTM and levosimendan treated groups (Fig. 5f, g). In levosimendan-treated group, AT8-immunoreactivity decreased to 0.4±0.1-fold in cortex and 0.4±0.2-fold in hippocampus, compared to those of vehicle-treated group. In LMTM-treated group, AT8-immunoreactivity decreased to 0.6±0.2-fold in cortex and 0.5±0.2-fold in hippocampus.
Next, tau immunoblot assay was followed to evaluate drug effects on tau oligomerization and aggregation, using RIPA-soluble and RIPA-insoluble brain lysates. On tau immunoblots, mouse tau is indicated with black arrow (50 kDa) and human tau-BiFC compartments are indicated with two green arrows (76 and 85 kDa). In RIPA soluble brain lysates, tau phosphorylation levels were signi cantly decreased in LMTM-and levosimendan-treated groups (Fig. 5h). In levosimendan-treated group, the levels of tau phosphorylation decreased 0.5±0.0-fold at S199 and 0.5±0.1-fold at S396, compared to that of vehicle-treated group. In LMTM-treated group, the levels of tau phosphorylation decreased 0.6±0.1-fold at S199 and 0.8±0.2-fold at S396. Different from the previous results showing the increased level of SDSand b-mercaptoethanol resistant tau oligomers, LMTM-treated group did not show any high-molecular weight oligomers in the brain lysates (Fig. 5h). Also, the level of total tau decreased slightly in levosimendan treated group, although the decrease was not statically signi cant. Corresponding to AT8immuno uorescence images in Figure 5f, the amount of insoluble tau decreased 0.5±0.2-fold in levosimendan-treated mice (Fig. 5i). In case of LMTM, the amount of insoluble tau decreased 0.7±0.1fold. Collectively, our results indicated that levosimendan is a new drug candidate targeting tau oligomerization, suppressing tau phosphorylation and aggregation in Tau P301L -BiFC mice. Corresponding to AT8-immuno uorescence images in Figure 5f, the amount of insoluble tau decreased 0.5±0.2-fold in levosimendan-treated mice (Fig. 5i). In case of LMTM, the amount of insoluble tau decreased 0.7±0.1fold.
Next, the level of disul de-linked tau oligomers was evaluated by non-reducing SDS-PAGE analysis. Similar to the result of primary neuron cultures, disul de-linked tau oligomers were observed in the brain lysates of wild type mice (Supplementary Fig. 4). In Tau P301L -BiFC mice, most of human and murine tau exists as the form of disul de-linked oligomers (> 245 kDa) (Fig. 5j and Supplementary Fig. 4). Levosimendan-treatment signi cantly decreased the high-molecular weight oligomers and increased soluble monomeric tau level. This result is consistent with previous data showing that levosimendan inhibits the formation of disul de-cross linkage between tau. In contrast, LMTM-treatment did not increase nor decrease the level of disul de-linked tau oligomers in the brain (Fig. 5j). The LMTM effect in the animal was contrasted with the effect when LMTM interacts with tau directly in test tubes. This result implies that when intraperitoneally administrated to mice, LMTM might not have direct interaction with tau as shown in vitro test (Fig. 3a, b). The anti-tau activity of LMTM shown in the brain of Tau P301L -BiFC mice could be the result of indirect regulation of tau pathway, such as decreasing tau phosphorylation or increasing autophagic clearance of aggregated proteins 7,46 . Collectively, our results indicated that levosimendan is a new drug candidate targeting tau oligomerization, suppressing tau phosphorylation and aggregation in Tau P301L -BiFC mice.

Discussion
Levosimendan was identi ed 20 year ago and has been treated to patients with acute heart failure 47 . As an inodilator, levosimendan possesses both positive inotropic and vasodilator actions; (i) as a calcium sensitizer, it enhances the sensitivity of contractile proteins to calcium through covalent binding to troponin C 48 . (ii) as a vasodilator, levosimendan inhibits phosphodiesterase III and opens ATP-dependent K+ channels in smooth muscle cells, leading to arteriolar and venous dilation 49 . In addition to heart, ATPdependent K+ channels present in a number of tissues including brain 50 . In 2010, Roehl et al reported neuroprotective effect of levosimendan in vitro model of traumatic brain injury, which might be associated with the activation of neuronal ATP-dependent K+ channels 51,52 . There are also evidences indicating that the neurorotective effect of levosimendan is associated with its action as a vasodilator. In 2015, Levijoki et al reported that orally administered levosimendan increased blood volume of the cerebral vessels, reducing mortality and morbidity in rat models of primary and secondary stroke 53  Here our study shows a new molecular mechanism of levosimendan as an anti-tau agent. By modifying tau cysteine residues, levosimendan inhibits disul de-linked tau oligomerization. Different from LMTM, which showed con icting effects on tau, levosimendan exhibited consistent, inhibitory effect on tau oligomerization in vitro, in cells, and in the brain of mice, suppressing tau aggregation. In tau-BiFC mice, 4 month-administration of levosimendan not only suppressed tau pathology, but also prevented memory de cits in aged Tau P301L -BiFC mice. To apply levosimendan as AD therapeutics, safety issue should be clari ed, due to its action as an inodilator. There is strong association between dementia and cardiovascular disease. Cardiovascular insu ciency impairs the function of diverse organs, including the brain, which can worsen a pathology related to dementia 55,56 . For an example, phosphodiesterase III is known to be upregulated in cerebral blood vessels of AD patients due to vascular amyloid burden, and phosphodiesterase III inhibitors have shown protective effects in AD model 57,58 . In that case, the vasodilator-action of levosimendan would be bene cial in improving AD brain function.
Drug repositioning is an attractive drug discovery strategy, which generates new therapeutic value from existing drugs 59 . MB is one of the representative cases of drug repositioning. In 1891, MB was rstly approved to treat malaria with its activity as a chloroquine sensitizer 60 . Then, MB had been approved to treat clinical pain syndromes, psychotic disorders, cyanide poisoning and urinary tract infections 61 . In 2016, MB was approved to treat Methemoglobinemia 62 . In addition to the known medical actions of MB, many studies have compiled wide variety of biological activities of MB, which i) inhibits the activity of monoamine oxidase A 63 , nitric oxide synthase 64 , and guanylate cyclase 65 , ii) increases the release of neurotransmitters, such as serotonin and norepinephrine 66,67 , iii) increases cholinergic transmission 68 , iv) inhibits GSK3b and microtubule-a nity regulating kinases 69,70 , and v) promotes autophagic clearance of b-amyloid 71 and tau 72 . The action of methylene blue on multiple targets in the brain justi es its symptom relieving effects on Tau P301L -BiFC mice; improving learning and cognitive abilities and reducing tau phosphorylation and aggregates. However, our in vitro and cell-based data clearly demonstrated that tau oligomer is not the direct therapeutic target of MB or LMTM.
In AD patients, tau oligomers are detected at the early stages of pathogenic cascades 14,15,73 . Therefore, preventing tau oligomerization is an important therapeutic strategy to prevent neuronal loss and memory de cits in AD patients. However, it has been di cult to establish therapeutic strategy preventing tau oligomerization, since tau pathology is linked with a number of cellular processes, which are closely linked to each other. For example, if GSK3b a major tau kinase, is inhibited, other tau kinases activate tau pathology 5,74 . Therefore, direct modi cation of tau would be more effective a strategy to prevent tau oligomerization without altering other cellular processes. Our results showed that levosimendan covalently binds to tau cysteine residues and inhibits tau oligomerization preventing tau pathology.
Moreover, levosimendan could disassemble tau-tau interactions regardless of its aggregation states. The binding mode between levosimendan and tau should be clari ed in future study; how levosimendan dissociates disul de-linked tau oligomers into monomers; whether levosimendan could distinguish cytosolic tau from microtubule-bound tau. Once levosimendan disassembles tau oligomers into monomers, the monomeric tau would be degraded by proteasome complexes, reducing soluble tau burden 75 . Our results supported this by showing the decreased level of tau in levosimendan-treated tau-BiFC cells and primary neurons, together with the reduced level of tau oligomers. Levosimendan displayed robust potency against tau oligomerization and rescued tauopathy-induced cognitive declines in Tau P301L -BiFC mouse model. Although careful validation is required, our data present the potential of levosimendan as a disease modifying drug for AD.

Materials And Methods
Source of chemicals FDA-approved & Passed Phase I Drug Library was purchased from Selleckchem. Forskolin and MB were purchased from Sigma, and leucomethylene blue mesylate (LMTM) was purchased from MedChemExpress. Levosimendan and its metabolites, OR-1855 and OR-1896 were purchased from Toronto Research Chemicals Inc. Synthesis and characterization of 14 C-levosimendan were conducted by Curachem, Inc. according to previously described protocol 76 . The characterization of 14 C-levosimendan is provided in in the Supplementary Materials.

Preparation of tauK18 fragments
TauK18 and tauK18 P301L fragments were expressed and puri ed according to previously described protocols 19,77 . 6xHis-tagged tauK18 and tauK18 P301L were expressed in E.coli BL21 (DE3) and puri ed by

Flow cytometry analysis
Tau-BiFC cells, grown in 6-well plates, were treated with MB, LMTM (5 µM), or levosimendan (15 µM), followed by the treatment of tauK18 P301L (5 µg/mL). After 24 hrs, BiFC cells were collected and subjected to ow cytometry analysis, using BD FACSLyric TM cytometer (BD Biosciences). A total of 50,000 events were acquired per sample. All samples were gated using the same gating tree and gate positions; side scatter area (SSC-A) vs. forward scatter area (FSC-A) for viable and singlet cell populations. BiFC uorescence was excited by 488 nm wave laser and collected through a 527/32 band-pass lter.

Tau-immunoblot analysis of reducing and non-reducing SDS-PAGE gels
For reducing SDS-PAGE analysis, cell or brain lysates were mixed with 4x Laemmli buffer containing 10% β-mercaptoethanol (BME) and boiled at 97℃ for 5 min. For non-reducing SDS-PAGE analysis, cell or brain lysates were mixed with 4x Laemmli buffer without BME. For immunoblot analysis, 10 μg of each lysate was separated into 10% SDS-PAGE gel and transferred to PVDF membrane. The levels of total tau and phosphorylated tau were detected by anti-tau antibody; 2B11 (IBL), Tau5 (Abcam), pSer199 (Abcam), pSer396 (Abcam), and pThr205 (Abcam). Band intensity was quantified using Image J software (NIH).
Primary neuron culture and preparation of the lysates Primary hippocampal neurons were isolated from day 18 embryonic Sprague-Dawley rat brains as described previously 78 . The neurons were seeded at a density of 3.5×10 5 cells/well on a poly-D-lysinecoated 6-well plate and maintained in the neurobasal medium at 37°C in a humidi ed atmosphere of 5% CO 2 . The neurobasal medium contains 2% B27 supplement, 0.5 mM glutamax, 100 units/mL penicillin, and 100 μg/mL streptomycin. Every 3 days, 50% of the medium was replaced with the fresh neurobasal medium. At DIV10, neurons were incubated with MB, LMTM (3 µM), or levosimendan (10 µM) in the presence of tauK18 P301L (10 µg/mL) for 48 hrs. Then, the neurons were washed with PBS and lysed in CelLytic M lysis reagent (Sigma) containing protease/phosphatase inhibitor cocktail (Sigma).

TauK18 oligomerization assay in vitro
To induce tau oligomerization, tauK18 (0.5 mg/mL dissolved in PBS, pH7.4) was incubated with each drug in the presence of DTT (100 μM) and heparin (0.1 mg/mL) at RT for 5 hrs with vigorous shaking.
Then, tau oligomers were separated on 4-20% SDS-PAGE gels under reducing and non-reducing condition and visualized by Coomassie blue stain.
In vitro tau aggregation and disaggregation assays.
To evaluate the inhibitory effect of drugs on tau aggregation, tauK18 protein (0.5 mg/mL) was incubated with each drug at various concentration in the presence of DTT (100 μM) and heparin (0.1 mg/mL) at 37℃ for 5 days, with vigorous shaking (220 rpm). For tau disaggregation assay, the preformed aggregation-mixture of tauK18 was incubated with each drug at 37℃ for 4 days, with vigorous shaking (220 rpm). At the nal day of the incubation, the level of b-sheet aggregates was evaluated by thio avin S (ThS) assay. For ThS assay, 5 μL of each mixture was transferred to a black 384-well plate with 45μL of PBS containing 10 μM ThS. ThS uorescence (λ ex = 430 nm, and λ em = 500 nm) was measured by using a Flexstation2 spectrophotometer (Molecular Devices).

Transmission electron microscopy (TEM)
Samples were placed onto carbon-coated copper electron microscopy grids, and then negatively stained with 2 % (w/v) aqueous uranyl acetate for 1 min. For imaging of tau laments (Fig. 3D), the grids were observed using a JEM-1011 transmission electron microscope (JEOL) at the acceleration voltage of 80 kV. For imaging of tau laments and oligomers ( Fig. 3F and Fig. 4E), the grids were observed using Tecni G2 F20 transmission electron microscope (FEI) at the acceleration voltage of 120 kV. Eight to twelve random images from each experimental condition were captured by the operator as a blind observer.
Autoradiography of 14 C-levosimendan TauK18 (1mg/mL, 72 μM) was incubated with 14 C-levosimendan (720 μM) in PBS containing 5% of DMSO at RT for 2 hrs. Then, the mixture was separated on 15% SDS-PAGE gel under non-reducing condition and stained with Coomassie blue. For autoradiography, the SDS-PAGE gel was transferred to PVDF membrane and scanned by Typhoon FLA 7000 IP (GE Healthcare).

MALDI-TOF analysis of R1-R4 peptides
Each repeat domain (R1-R4, 150 mM) was incubated with levosimendan (molar ratio 1:10 of a tau repeat domain: levosimendan) at RT for 37 hrs. The samples were mixed with matrix solution (10 mg/mL of Sinapinic acid in 0.1 % (v/v) tri uoroacetic acid/CAN) at a ratio 1:1. The mixtures were directly spotted onto the MALDI target and dried. Mass spectra were acquired in re ection/linear positive ion mode in the m/z range of 2,000-100,000 using an Ultra ex III TOF/TOF mass spectrometer controlled by Flex Control 3.0 (Bruker Daltonics). Default operating conditions are as follows: ion source 1, 25.0 kV; ion source 2, 23.0 kV; lens voltage, 6.0 kV; laser repetition rate = 100 Hz. All spectra were generated automatically in the instrument software and based on averaging 1000 shots from 10 non-overlapping positions (100 shots/position).
Animal studies Tau P301L -BiFC mice Tau P301L -BiFC mice were bred and maintained as described previously 33 . Tau P301L -BiFC mice were bred with C57BL/6N mice and maintained in pathogen-free facilities. Their heterogeneous offspring and wildtype littermates were used in this study. All mice were allocated randomly for experiments, but groups were counterbalanced for animal sex and group average body weight. Animal protocols followed the principles and practices outlined in the approved guidelines by the Institutional Animal Care and Use Committee of the Korea Institute of Science and Technology. All animal experiments were approved by the Korea Institute of Science and Technology.
Drug administration to Tau P301L -BiFC mice For drug administration, LMTM and levosimendan were dissolved in PBS containing 40% polyethylene glycol (PEG; Sigma). 9 month-old Tau P301L -BiFC mice were intraperitoneally administered with each drug for 4 months, three times a week, with 5 mg/kg dosage (n=11 per group). Vehicle-treated Tau P301L -BiFC mice and age-matched wild type mice were used as control groups. Behavioral assessments were conducted as described below for one month from the end of drug administration. All mice were sacri ced at the age of 14 months for pathological analysis.

Behavioral tests
Novel object recognition. Novel object recognition test was performed as described previously 79 . Brie y, for the habituation, mice (n=9 per group) were individually placed in the center of an open eld arena (40 × 40 cm) and allowed to freely explore for 15 min. Next day, in the training trial (familiarization phase), mice were allowed to explore two identical objects in the open eld for 10 min. In the testing trial (recognition phase) performed 24 hours later, one familiar object was changed for a novel object which was different in color and shape. Mice were allowed to explore the objects for 10 min. The exploration time for the familiar (old) or the new (novel) object during the recognition phase was recorded using Noldus EthoVision XT video tracking system. Memory was operationally de ned by the recognition index calculated by dividing the time an animal spent exploring the novel object or old object by the total time spent exploring objects in the testing period. The exploration time was measured when the mouse was pointing towards the object in the vicinity of the object. Y-maze. Mice (n=11 per group) were tested for spontaneous alternation behavior in a Y-shaped maze (60 cm in length of each arm, 20 cm in depth) using standard protocol 80 . Spontaneous alternation for 8 mins was calculated as the proportion of alternations. Alteration (%) was de ned as consecutive entries in three different arms (ABC), divided by the number of possible alterations (total arm entries minus 2) 81 .
Passive avoidance. Passive avoidance test was performed as described previously 82 with the following modi cations. Mice (n=8 per group) were adapted in the passive avoidance chamber (Gemini) for 10 min, and then were returned to their home cages. The chamber is composed of a light compartment and a dark compartment separated by a connecting gate. The following day, mice were placed in the light compartment, and the gate was opened after 30 seconds. When mice entered the dark compartment, the gate was closed and an electrical foot shock (3 mA) was delivered for 2 sec. Mice were left in the dark compartment for 30 sec after the foot shock to enable association of the environment with the aversive stimulus. Mice were then returned to their home cages. The following day, mice were placed in the light compartment again, and the gate was opened after 30 seconds. The step-through latency, the time required for mice to enter the dark compartment, was measured up to 540 sec.

Preparation of brain tissue slices
Mice were perfused with 0.9% saline and xed with PBS (pH 7.4) containing 4% paraformaldehyde (Sigma). Brains were extracted and xed in PBS containing 4% paraformaldehyde at 4°C for 48~72 hrs.
For cryoprotection, the brains were transferred to PBS containing 30% sucrose solution and incubated at 4°C until they sunk. For cryosectioning, the brains were embedded with O.C.T (Tissue-TEK), and cut serially using a cryostat (CM1860UV, Leica). 30-μm thick tissue slices were maintained in PBS containing 0.05% sodium azide at 4°C.

Sudan Black B stain and BiFC uorescence imaging
Brain tissue section (n=8 per group) were mounted onto slides. To reduce auto uorescence, brain tissue sections were stained with Sudan Black B solution (70% ethanol containing 0.05% Sudan Black B) for 10 min, washed three times with PBS containing 0.1% Triton X-100 (Sigma), and then washed with distilled water. For counter-staining of nuclei, brain tissues were stained with Hoechst (0.5 μg/mL) for 30 min. BiFC uorescence (λ ex = 460-490 nm and λ em = 500-550 nm) images were acquired using Axio Scan. Z1 (ZEISS). For quantitative analysis, Mean BiFC-uorescence intensity was measured in somatosensory cortex (layer V) and hippocampal CA1 from AP;-1.82. To normalize the background uorescence between the brain samples, habenular region was used as an internal control. Using image J software (NIH), the region of interest was masked, and the uorescence intensity value was calculated. Data were shown as the average and standard deviation of Mean uorescence intensities of brain images from six to eight animals.

Preparation of RIPA-soluble and insoluble brain lysates
Brains were weighed and suspended in RIPA lysis buffer (Sigma) containing protease and phosphatase inhibitor cocktails. Then, tissues were mechanically disrupted using a cordless mortar and pestle (Sigma), and incubated at 4°C for 2 hrs, with shaking. The homogenates were centrifuged at 20,000 g at 4°C for 20 min. The supernatants were collected as RIPA-soluble fractions and stored at −80°C. To prepare RIPAinsoluble fractions, the remaining pellets were washed once with RIPA lysis buffer containing 1 M sucrose, and then resuspended in 2% SDS solution (1 mL per gram of tissue) and incubated at RT for 1 hr. The mixtures were centrifuged at 20,000 g for 1 min at RT. The supernatant was collected as RIPAinsoluble fractions and preserved at −80 °C.

Statistics
All data in quantitative analysis were presented as mean ± S.D. or S.E.M.. Unpaired t-test was performed when two groups were compared. One-way ANOVA or two-way ANOVA was performed when multiple groups were compared, depending on the number of independent variables. Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software).    Levosimendan inhibits disul de-linked tau oligomerization by capping tau cysteine residues a, c Schematic diagram of a thioimidate bond formation of 14C-levosimendan and tau cysteine (a). TauK18 pre-aggregates were incubated with DMSO control or 14C-levosimendan for 2 hrs at RT. The mixtures were separated on a non-reducing SDS-PAGE gel (15%) for further analysis. Coomassie blue stain indicates disul de-linked tau oligomers on the SDS-PAGE gel (b). The SDS-PAGE gel was transferred onto