AQAMAN, a bisamidine-based inhibitor of toxic protein inclusions in neurons, ameliorates cytotoxicity in polyglutamine disease models

Polyglutamine (polyQ) diseases are a group of dominantly inherited neurodegenerative disorders caused by the expansion of an unstable CAG repeat in the coding region of the affected genes. Hallmarks of polyQ diseases include the accumulation of misfolded protein aggregates, leading to neuronal degeneration and cell death. PolyQ diseases are currently incurable, highlighting the urgent need for approaches that inhibit the formation of disaggregate cytotoxic polyQ protein inclusions. Here, we screened for bisamidine-based inhibitors that can inhibit neuronal polyQ protein inclusions. We demonstrated that one inhibitor, AQAMAN, prevents polyQ protein aggregation and promotes de-aggregation of self-assembled polyQ proteins in several models of polyQ diseases. Using immunocytochemistry, we found that AQAMAN significantly reduces polyQ protein aggregation and specifically suppresses polyQ protein–induced cell death. Using a recombinant and purified polyQ protein (thioredoxin–Huntingtin–Q46), we further demonstrated that AQAMAN interferes with polyQ self-assembly, preventing polyQ aggregation, and dissociates preformed polyQ aggregates in a cell-free system. Remarkably, AQAMAN feeding of Drosophila expressing expanded polyQ disease protein suppresses polyQ-induced neurodegeneration in vivo. In addition, using inhibitors and activators of the autophagy pathway, we demonstrated that AQAMAN's cytoprotective effect against polyQ toxicity is autophagy-dependent. In summary, we have identified AQAMAN as a potential therapeutic for combating polyQ protein toxicity in polyQ diseases. Our findings further highlight the importance of the autophagy pathway in clearing harmful polyQ proteins.

in long-term activation of autophagy and the UPR pathway in neurons, which are likely compensatory mechanisms to relieve the ER stress (14,15). The disruption of UPR or autophagy causes inefficient clearance of the accumulated proteins, which in turn contributes to the progression of neurodegeneration and cell death (13,14).
No cure for polyQ diseases is known, so new therapeutics aimed at disaggregating or inhibiting the formation of the abnormal polyQ protein inclusions are urgently needed. Bisamidine-based compounds are a type of soluble small molecules capable of binding nucleic acids (16,17). In this study, we found that a bisamidine inhibitor, AQAMAN, can reduce polyQ protein aggregation and suppress polyQ protein-induced cell death. We demonstrated that AQAMAN can bind to soluble forms of polyQ proteins and disrupt preformed polyQ aggregation in vitro. Notably, feeding AQAMAN to Drosophila with expanded polyQ protein expression suppressed neurodegeneration in vivo. AQAMAN also suppressed ER stress in both cell culture and Drosophila models of polyQ diseases. We further demonstrated that AQAMAN's protective effect on cells is autophagic pathway-dependent.
In summary, AQAMAN is a potent bisamidine inhibitor that effectively reduces polyQ protein aggregation, which has significant therapeutic potentials for combating polyQ diseases.

AQAMAN suppresses polyQ protein-induced cell death
Our previous work has demonstrated that by using rational design, a series of bisamidine inhibitors can be designed to bind to the RNA groove of CUG repeats and to alleviate RNA toxicity in myotonic dystrophy type 1 (18). To search for small molecules that can ameliorate cellular toxicity in polyQ diseases, we screened through a library of bisamidine inhibitors, and identified N 1 ,N 3bis(2-(2,4,6-triaminopyrimidin-5-yl)ethyl)isophthalimidamide (Fig. 1A). This molecule contains one aromatic ring and two pyrimidine moieties and was originally designed for inhibiting the formation of the MBNL1-(CCUG) exp protein-RNA complex in myotonic dystrophy type 2 (19). We termed this compound as Anti-polyQ Aggregation for Machado-Joseph-Associated Neurodegeneration (AQAMAN). We first examined whether AQAMAN induces any detectable cytotoxicity. Up to 100 M AQAMAN induced no detectable cell death in rat primary cortical neurons (Fig. S1A) and SK-N-MC cells (Fig. S1B).
To investigate AQAMAN's neutralizing effect on polyQ expansion-induced cytotoxicity, we utilized an established model of polyQ diseases by transfecting SK-N-MC cells with EGFP CAG81(R ϩ P) (expressing both expanded CAG RNA and polyQ protein) (20), and we assessed the levels of cell death using an established lactate dehydrogenase (LDH) cytotoxicity assay (21). Cells expressing the pathogenic EGFP CAG81(R ϩ P) displayed significantly higher levels of cell death compared with the control cells expressing EGFP CAG19(R ϩ P) (Fig. 1B). We found that the application of 0.5 M AQAMAN partially suppressed the EGFP CAG81(R ϩ P) -induced cell death (Fig. 1B). Increasing the concentration of AQAMAN to 1.0 or 2.0 M had no further suppression on cell death, indicating that the suppressive effect had already reached saturation at 0.5 M in this model (Fig. 1B). To determine whether AQAMAN suppressed cell death by neutralizing RNA toxicity and/or protein toxicity, we utilized an expanded CAG RNA toxicity model of polyQ diseases by expressing EGFP CAG78(R) (RNA toxicity) in SK-N-MC cells (Fig. 1C) (22). No suppression of cell death was detected upon the treatment of cells with up to 2.0 M AQAMAN, indicating that AQAMAN has no effect on neutralizing expanded CAG repeat-induced RNA toxicity (Fig. 1C). To ensure that the cell death suppression effect by AQAMAN is consistent in other models of polyQ diseases, we further utilized an established cell model of MJD. In this model, SK-N-MC cells expressing truncated MJD CAG78(R ϩ P) (trMJD CAG78(R ϩ P) ) possess both CAG RNA toxicity and polyQ protein toxicity, and they display significantly higher amounts of cell death compared with control cells expressing the control unexpanded trMJD CAG27(R ϩ P) (Fig. 1D) (22). Although the treatment of cells with 0.5 M AQAMAN had no suppressive effect on this model, 1.0 and 2.0 M AQAMAN treatment significantly suppressed cell death (Fig. 1D). Apart from the codon CAG, CAA also encodes for glutamine. Thus, in a construct that carries interrupted CAA/CAG repeats, the RNA toxicity component is disrupted while protein toxicity is retained (22,23). Similar to cells expressing trMJD CAG78(R ϩ P) , cells expressing trMJD CAA/G78(P) (protein toxicity) exhibited an increased amount of cell death (Fig. 1E). Application of 0.5 M AQAMAN in this model was sufficient to achieve a robust suppression of cell death, indicating that AQAMAN suppressed polyQ protein-induced cell death (Fig. 1E). Increasing the concentration of AQAMAN to 1.0 or 2.0 M had no further effect, indicating that AQAMAN's suppression on polyQ protein toxicity was already saturated at 0.5 M (Fig. 1E).

AQAMAN reduces the aggregation of polyQ proteins
To determine how AQAMAN suppresses polyQ proteininduced cell death, we transfected SK-N-MC cells with EGFP CAG81(R ϩ P) and performed immunocytochemistry (ICC) to visualize polyQ aggregation with or without AQAMAN treatment. We found that 1.0 M AQAMAN effectively reduced polyQ-containing EGFP aggregation in the EGFP CAG81(R ϩ P)transfected SK-N-MC cells (Fig. 2, A and B). Consistently, 1.0 M AQAMAN also effectively reduced polyQ-containing aggregation in SK-N-MC cells expressing EGFP-tagged trMJD CAG78(R ϩ P) (Fig. 2, C and D). Importantly, similar protective effects of AQAMAN were also observed in rat cortical neurons transfected with EGFP CAG81(R ϩ P) (Fig. 2, E and F) or EGFP-tagged trMJD CAG78(R ϩ P) (Fig. 2, G and H), demonstrating the neuroprotective effect of AQAMAN.
In addition to ICC, filter trap and Western blotting analyses were performed to assess the effectiveness of AQAMAN in reducing polyQ protein aggregation. In our filter trap assay, we showed that 0.5 M AQAMAN treatment resulted in a detectable reduction of EGFP-Q81 protein aggregation (Fig. 3A). Consistent with our results in the cell death assay shown in Fig.  1B, increasing the concentration of AQAMAN to 1.0 or 2.0 M had no further anti-aggregation effect (Fig. 3A), indicating that saturation was reached at 0.5 M in this model. Meanwhile, Western blot analysis showed that, upon AQAMAN treatment, soluble EGFP-Q81 level increased (Fig. 3B). To ensure that the

Bisamidine-based inhibitor reduces polyQ aggregation
increase of soluble polyQ proteins was not due to altered total protein expression, we performed formic acid treatment on the samples to solubilize the polyQ proteins (24). Our results showed that the total amount of protein remained unchanged (Fig. 3C). We then further tested AQAMAN's anti-aggregation effect on our MJD model of trMJD CAG78(R ϩ P) expression. Consistent with our findings in the cell death assay in Fig. 1D, our results showed that 1.0 M AQAMAN was sufficient to reduce trMJD-Q78 aggregation (Fig. 3D). In Western blot analysis, consistent with previous studies, the soluble trMJD-Q78 proteins were shown as a doublet band (25,26). An increase in the soluble trMJD-Q78 proteins was observed when AQAMAN was applied ( Fig. 3E). When we solubilized all proteins using formic acid, no detectable change in the total amount of protein expressed was found (Fig. 3F), suggesting that AQAMAN treatment increased the amount of soluble trMJD-Q78 in Fig. 3E.

Anti-aggregation effect of AQAMAN depends on its pyrimidine pendants
To determine whether AQAMAN prevents the formation of polyQ aggregates and/or breaks down preformed polyQ aggregates, we expressed the polyQ protein, Trx-Htt-Q46, in Escherichia coli and allowed the purified polyQ proteins to aggregate in a cell-free system by cleaving off the thioredoxin tag (27).

Bisamidine-based inhibitor reduces polyQ aggregation
We then immediately applied AQAMAN (0-h pre-aggregation) to test its ability in preventing aggregate formation. Our results showed that 50 M AQAMAN was sufficient to significantly reduce the formation of polyQ aggregates (Fig. 4A), and 500 M AQAMAN was even more effective in reducing aggregate formation (Fig. 4A). Thereafter, to examine whether AQAMAN can break down preformed polyQ aggregates, we applied AQAMAN after a 24-h pre-aggregation period. Our results showed that 50 M AQAMAN was sufficient to break down the preformed aggregates in vitro (Fig. 4B), and 500 M AQAMAN showed an even more prominent effect of breaking down preformed aggregates (Fig. 4B).

Bisamidine-based inhibitor reduces polyQ aggregation
To determine what structural features of AQAMAN contribute to its anti-aggregation effect, we synthesized another bisamidine control compound, AMD1. The general chemical features of AQAMAN and AMD1 are similar, both compounds are bisamidines with two aromatic pendants linked to the central aromatic core via a short alkyl linker, except that the pyrimidine pendants of AQAMAN were replaced by phenol units (Fig. 4C). AMD1 was unable to prevent the formation of polyQ aggregates (Fig. 4D) nor was it able to break down preformed polyQ aggregates (Fig. 4E). This finding suggests that the ability of AQAMAN to deaggregate polyQ proteins depends on its pyrimidine pendants.

AQAMAN mitigates neurodegeneration in a Drosophila model of MJD
To investigate whether AQAMAN can mitigate polyQ protein toxicity in vivo, we employed an established Drosophila model of MJD by expressing full-length MJD CAG84 (flMJD CAG84 ) in the fly eye using the gmr-GAL4 driver to induce retinal degeneration (28). Flies expressing flMJD CAG84 showed a significant reduction of rhabdomeres per ommatidium compared with control flies that were expressing the unexpanded flMJD CAG27 (Fig. 5, A and B). flMJD CAG84 flies fed either 40 or 80 M AQAMAN resulted in partial rescue of retinal degeneration (Fig. 5, A and B). Results of the filter trap and Western blot analyses demonstrated that 40 and 80 M AQAMAN treatment effectively reduced polyQ aggregation in flies (Fig. 5, C and D). Formic acid treatment was used to verify that the total flMJD-Q84 protein level was unchanged (Fig. 5E). To confirm that AQAMAN has no effect on RNA toxicity in vivo, we examined AQAMAN's effect on a RNA toxicity-only model by expressing DsRed CAG100 in the fly eye (23). In this transgene, the CAG repeat was placed at the 3Ј UTR of the DsRed reporter, and thus, no polyQ protein would be produced. As expected, feeding the flies with 80 M AQAMAN was unable to rescue the DsRed CAG100 -induced retinal degenera- showing that the total trMJD-Q78 protein levels did not change in response to AQAMAN treatment. D-F, anti-HA was used to detect trMJD-Q27 and trMJD-Q78, and anti-␤-tubulin was used for loading controls. Error bars represent S.E. All experiments were performed at least three times independently.

Bisamidine-based inhibitor reduces polyQ aggregation
tion caused by the RNA toxicity of the untranslated CAG repeats (Fig. 5, F and G). These results demonstrated that AQAMAN can ameliorate neurodegeneration in vivo by reducing the aggregation of polyQ proteins.

Neutralization of polyQ protein toxicity by AQAMAN requires autophagy
Autophagy is a tightly-regulated catabolic mechanism for lysosomal degradation of misfolded proteins (29). Impaired autophagy results in the accumulation of misfolded proteins, contributing to ER stress (13,14). Even though AQAMAN is capable of reducing the aggregation of polyQ proteins, we speculated that autophagy would remain necessary for their clearance to reduce toxicity. To address this, we either blocked or induced autophagy in our cell model of polyQ diseases and examined AQAMAN's effect on cell death using LDH assay and polyQ aggregation. Wortmannin is a phosphatidylinositol 3-kinase inhibitor that impedes autophagosome formation, thereby blocking autophagy (30), whereas rapamycin is a mammalian target of rapamycin (mTOR) inhibitor that mimics cellular starvation by blocking signals required for cell growth and proliferation, which in turn activates autophagy (30). We found that blocking autophagy with wortmannin had no effect on the untransfected or the control unexpanded trMJD CAG27(R ϩ P)expressing cells but resulted in increased cell death in expanded polyQ trMJD CAG78(R ϩ P) -expressing cells (Fig. 6A). Consistent with our findings in Fig. 1D, AQAMAN treatment suppressed cell death in trMJD CAG78(R ϩ P) -expressing cells (Fig. 6A). However, under the effect of wortmannin, AQAMAN was unable to suppress cell death in trMJD CAG78(R ϩ P) -expressing cells (Fig. 6A). These results suggested that the suppression of polyQ-induced cell death by AQAMAN depends on the autophagic pathway.
It is possible that wortmannin blocked AQAMAN's cytoprotective effect by impeding its anti-aggregation function. Thus, we performed ICC to examine polyQ aggregation under the effect of these compounds. Consistent with our findings in Fig.  2, C and D, AQAMAN treatment reduced polyQ aggregates in trMJD CAG78(R ϩ P) -expressing cells (Fig. 6, B and C). However, the presence of wortmannin did not prevent AQAMAN from reducing polyQ aggregates (Fig. 6, B and C). These results suggest that although the suppression of polyQ-induced cell death by AQAMAN depends on autophagy, the anti-aggregation effect of AQAMAN is independent of autophagy.
Because blocking autophagy abolished AQAMAN's effect on suppressing cell death, we sought to determine whether the  (24 h). C, molecular structure of AQAMAN and AMD1. AMD1 has a similar structure to AQAMAN but with the pyrimidines (red) replaced by phenols (blue). D and E, cell-free pre-aggregation assay, showing that AMD1 has no effect on polyQ protein aggregation prevention (0 h) (D) or deaggregation of preformed inclusions (24 h) (E). Error bars represent S.E. All experiments were performed at least three times independently.

Bisamidine-based inhibitor reduces polyQ aggregation
activation of autophagy using rapamycin would enhance the cytoprotective effects of AQAMAN. Similar to the AQAMAN treatment, rapamycin treatment suppressed polyQ-induced cell death (Fig. 6D). When rapamycin was added to AQAMANtreated cells, it resulted in further suppression of cell death compared with cells treated with AQAMAN alone (Fig. 6D), indicating that activation of autophagy could indeed further potentiate the suppressive effect of AQAMAN on polyQ-induced cell death. As for polyQ aggregation, both AQAMAN and rapamycin were capable of individually reducing cellular aggregates (Fig. 6, E and F). Application of both compounds together resulted in a further reduction of aggregates (Fig. 6, E  and F). The combinatorial effect of the two compounds appeared to be additive.

AQAMAN relieves ER stress
In polyQ diseases, misfolded proteins with polyQ tracts accumulate in the ER and cause ER stress (4, 5, 7). We examined

Bisamidine-based inhibitor reduces polyQ aggregation
the transcript level of the ER stress sensor binding immunoglobulin protein (BiP). BiP is a molecular chaperone located in the ER lumen that regulates the translocation of proteins. The expression of BiP is up-regulated during ER stress, triggering the UPR pathway (31,32). It has been shown that BiP expression is induced in polyQ diseases (33). Using quantitative RT-PCR, we demonstrated that trMJD CAG78(R ϩ P) -expressing SK-N-MC cells have increased expression of BiP compared with the control trMJD CAG27(R ϩ P) -expressing cells (Fig. 7A). We found that 0.5-2.0 M AQAMAN effectively suppressed BiP induction in trMJD CAG78(R ϩ P) -expressing cells (Fig. 7A), suggesting that ER stress was relieved. To further investigate whether similar effects could be observed in vivo, we exam-ined whether AQAMAN treatment could reduce BiP transcript levels in the flMJD CAG84 Drosophila model. Consistently, 40 and 80 M AQAMAN restored BiP transcription level back to that of control (Fig. 7B). These results demonstrated that AQAMAN could relieve polyQ-induced ER stress in vitro and in vivo.

Discussion
The abnormal aggregation of proteins is a hallmark of many neurodegenerative diseases, such as Alzheimer's disease, Parkinson's disease, and polyQ diseases (34). Because mutant protein aggregates have been shown to exhibit cellular toxicity (35) and even have the potential to spread to

Bisamidine-based inhibitor reduces polyQ aggregation
other cells and brain regions (36), controlling protein aggregation is therefore paramount to combating these neurodegenerative diseases.
Polyglutamine aggregates are mainly composed of fibers with ␤-sheet structure, which is similar to amyloid in Alzheimer's disease (3,37,38). In this study, we identified AQAMAN as a novel anti-polyQ aggregation compound. We showed that AQAMAN is capable of suppressing cell death (Fig. 1, B-E) and neurodegeneration (Fig. 5) induced by polyQ protein toxicity. Moreover, we demonstrated that AQAMAN can prevent polyQ aggregation and dissociate preformed polyQ aggregates in vitro (Fig. 4), which suggests that AQAMAN can likely bind to the soluble forms of the polyQ proteins and shift the aggregation equilibrium toward the soluble forms of the proteins. This effect was absent with the structurally similar AMD1, indicating that this binding was likely via the two pyrimidine pendants of AQAMAN (Fig. 4).
Two major pathways are responsible for degrading misfolded proteins in the cell: the ubiquitin-proteasome pathway and the autophagy pathway (29). When proteins are accessible to both pathways, the ubiquitin-proteasome system is the prioritized clearance route due to its higher efficiency. However, if a protein is aggregate-prone and the proteasomes fail to degrade the protein aggregates, then the autophagy pathway will become the main clearance route (29). Autophagy was originally identified as a self-consuming cellular process and was later found to be involved in the degradation of misfolded proteins as a cytoprotective response especially during aging and neurodegeneration (39). The autophagy pathway plays a pivotal role in the degradation of polyQ proteins and protects the cell from polyQinduced neurodegeneration and cell death (40,41). Despite the fact that AQAMAN can deaggregate polyQ proteins (Figs. 3 and 4), it is possible that the deaggregated polyQ proteins still exhibit substantial cytotoxicity and require clearance by the autophagy pathway. In fact, it was reported that large protein aggregates may not be toxic and that the smaller microaggregates may be the actual major toxic species (42,43). Indeed, recent studies have revealed that even deaggregated polyQ proteins may also be toxic (44). This is consistent with our observation that blocking autophagy does not affect AQAMAN's ability to reduce polyQ aggregation (Fig. 6, B and C) but abolished its effect on suppressing cell death (Fig. 6A). These results suggest that the dissociation of insoluble aggregates does not prevent cell death. Even after AQAMAN has broken down the visible aggregates, there may still be toxic species in the cell, such as microaggregates or soluble small oligomers that can contribute to polyQ-induced cytotoxicity. Functional autophagic pathways are still required for the proper clearance of these toxic species.
Previous studies have shown that the autophagy activator rapamycin can protect against polyQ-induced toxicity (45), and its cytoprotective effect depends on the activation of autophagy for the clearance of polyQ proteins (46,47). In line with these studies, we demonstrated that increasing the level of autophagy with rapamycin results in further reduction of polyQ aggregate and cell death in AQAMAN-treated cells (Fig. 6, D-F).
The accumulation of unfolded/misfolded polyQ proteins causes ER stress (4, 5, 7). Clearance of unfolded/misfolded proteins in the ER primarily relies on the UPR pathway, which induces the expression of molecular chaperones and folding enzymes for the ER to cope with the deleterious effects of unfolded/misfolded proteins (10,11). Our results showed that AQAMAN ameliorates polyQ-induced ER stress (Fig. 7). It is possible that the level of polyQ proteins deaggregated by AQAMAN are being kept low in the cell by autophagy. This, in turn, reduces the accumulation of unfolded/misfolded polyQ proteins in the ER, thereby relieving ER stress.
To further examine polyQ deaggregation by AQAMAN, we performed isothermal titration calorimetry (ITC) to study whether AQAMAN binds to Trx-Htt-Q46. Preliminary ITC results indicate a weak interaction between AQAMAN and Trx-Htt-Q46 (Fig. S2), suggesting that AQAMAN may not bind to the protein monomer, but its ability to reduce polyQ aggregation may be due to its binding to other soluble forms of the protein. More importantly, the observations that AQAMAN can both prevent the formation and solubilize the insoluble polyQ aggregates (Fig. 4, A and B) suggest that the polyQ aggregation and precipitation equilibria are reversible with no obvious kinetic effect. The ability of AQAMAN to sol-

Bisamidine-based inhibitor reduces polyQ aggregation
ubilize polyQ aggregates could therefore be attributed to its ability to bind to the soluble forms of the protein, which shifts the aggregation equilibrium from the insoluble aggregate to the soluble forms. This binding of AQAMAN to soluble polyQ proteins depends on the two pyrimidine pendants in its chemical structure as we showed that AMD1 did not have this effect (Fig.  4, D and E). Given the complexity of the polyQ aggregation equilibria involving different soluble forms of the protein and the insoluble aggregates, more elaborated binding studies on de-convoluting the heat profile to individual equilibrium will be necessary for a more detailed mechanistic insight. Nevertheless, it is clear that AQAMAN can inhibit the formation of the polyQ aggregates and neutralize its toxicity.
Many potential therapeutics against polyQ diseases have been previously reported. These therapeutics include but are not limited to small molecules and peptide inhibitors. Calmidazolium chloride was previously reported to prevent expanded polyQ-containing Huntingtin exon1 (Httex1) from aggregating (48). However, unlike AQAMAN, calmidazolium chloride targets only the initial step of Httex1 aggregation. Other polyQ small molecule inhibitors include those that modulate insulin/insulin-like growth factor signaling (49), inhibit the Rho-associated kinase (50), Hsp90 (51), and topoisomerase 1 (52) or accelerate autophagy (53,54). Peptide inhibitors, such as QBP1 (55), can inhibit polyQ aggregation by impeding the ␤-sheet conformational transition of the polyQ protein monomer as well as oligomer formation (56,57), whereas other peptide inhibitors, such as P3, P3V8, and TAT-BIND, that can bind directly to CAG RNA suppress RNA toxicity (58 -60). Although these peptide-based drugs are effective in reducing polyQ-induced cytotoxicity, their large molecular size may hinder them from crossing the blood-brain barrier (BBB). By contrast, small molecules are likely to be more efficient in penetrating the BBB. To assess whether AQAMAN can penetrate the BBB, we employed the on-line Blood Brain Barrier Predictor developed by Xie and co-workers (61). AQAMAN is indeed predicted to be able to penetrate the BBB (Fig. S3).
In this study, we identified AQAMAN as a small molecule that can ameliorate polyQ-induced protein toxicity by deaggregating polyQ inclusions. However, one apparent limitation of AQAMAN is its risk in promoting the buildup of small toxic oligomers or microaggregates and its dependence on functional autophagic pathways to properly relieve the cell from polyQ toxicity. AQAMAN treatment under impaired autophagy may accelerate the accumulation of such toxic species and may actually be harmful to the cell. Nevertheless, AQAMAN is a potent polyQ aggregation inhibitor that can likely cross the BBB, and thus, it has great therapeutic potential in bringing relief to patients suffering from incurable protein aggregation diseases.

Summary
We demonstrated the effect of AQAMAN in alleviating protein toxicity in several models of polyQ diseases. In vitro, AQAMAN can deaggregate polyQ proteins and suppress cell death in an autophagic pathway-dependent manner, whereas in vivo, AQAMAN can relieve polyQ-induced ER stress and ame-liorate neurodegeneration. These results underpin the potentials of AQAMAN as a therapeutic for polyQ diseases. Our study also highlights the importance of the autophagy pathway in the clearance of deaggregated polyQ proteins.

Drosophila culture and drug treatment
The Drosophila strains UAS-flMJD CAG27 , UAS-flMJD CAG84 (28), UAS-DsRed CAG0 , and UAS-DsRed CAG100 (23) used in this study were from Professor Nancy Bonini (University of Pennsylvania). gmr-GAL4 was obtained from Bloomington Drosophila Stock Center. Drosophila strains were cultured with cornmeal yeast glucose agar medium and maintained at 22°C. Genetic crosses were also carried out at 22°C. AQAMAN solution was added into fresh medium, which was subsequently used for fly culture.

Drosophila pseudopupil assay
Details of the assay were previously described (6). In a typical procedure, fly heads were cut and observed under light microscopy (Olympus BX51, Tokyo, Japan) with a ϫ60 oil objective. Images of ommatidia were captured using the SPOT Advanced software (Version 4.1; Diagnostic Instruments Inc., Sterling Heights, MI). For quantification of ommatidia integrity, a total of 200 ommatidia from 20 eyes of 10 individual flies were examined in each condition. The average number of rhabdomeres per ommatidium was counted. Each experiment was repeated at least three times.

SK-N-MC cell culture and drug treatment
The human neuroblastoma cell line SK-N-MC cells were obtained from ATCC (Manassas, VA) and cultured using DMEM (Hyclone; ThermoFisher Scientific, Waltham, MA) supplemented with 4 mM L-glutamine, 10% fetal bovine serum, and 1% penicillin-streptomycin. The cells were maintained at 37°C under 95% air, 5% CO 2 . For cells subjected to LDH assays, phenol red-free DMEM (Life Technologies, Inc., and ThermoFisher Scientific, Waltham, MA) was used for culturing. The AQAMAN solution was added into the medium at the same time of transfection and the treatment lasted for 72 h.

Transfection conditions
For the transfection of SK-N-MC cells, the cells were seeded at a density of 0.8 ϫ 10 5 cells/24-well plate. After 24 h, 1 g of plasmid DNA was used to transfect cells using Lipofectamine 2000 reagent (Invitrogen; ThermoFisher Scientific). For the transfection of primary rat cortical neurons, the neurons were incubated with 3 g of plasmids in primary neuron transfection reagent (VVPG-1003, Lonza Bioscience), followed by transfection using Amaxa Nucleofection system (Lonza Bioscience) according to the manufacturer's instructions.

Lactate dehydrogenase cytotoxicity assay
Details of the assay were previously described (21,60). Cyto-Tox 96 nonradioactive cytotoxicity assay kit (Promega, Madison, WI) was used to study cell viability. Both of the LDH released from culture medium and adherent cells were obtained, representing dead and survival cells, respectively. The percentage of cell death was calculated and normalized to the untransfected control.

Immunofluorescence staining
SK-N-MC cells were seeded on coverslips in 24-well plates. The culture medium was discarded, and cells were fixed with 3.7% formaldehyde fixative solution and permeabilized with 0.1% Triton X-100. After blocking with 5% goat serum, the cell nuclei were stained with 5 M Hoechst 33342 solution (Invitrogen; ThermoFisher Scientific). For the immunostaining performed on primary rat cortical neurons, the neurons were cultured at 12-well plates with coverslips and pre-coated with poly-D-lysine (Sigma). Confocal microscopy was performed on Leica TCS SP8 with software control of LAS X (Leica Microsystems CMS GmbH, Mannheim, Germany). Image analysis was performed using ImageJ (National Institutes of Health, Bethesda, MD) and Adobe Photoshop 7.0 software (Adobe, San Jose, CA). For quantification analysis, the percentage of cells that contain aggregated protein in transfected cells was calculated. For each group, over 200 cells were counted.

Quantitative real-time PCR (qRT-PCR)
A total of 10 adult fly heads or cultured cell samples (in standard 24-well plates) were homogenized in TRIzol reagent for extracting RNA. One microgram of RNA was then subjected to reverse transcription using the ImPromII TM reverse transcription system (Promega, Madison, WI). TaqMan probe qRT-PCR was applied to detect mRNA expression level. Quantification of target gene expression level was calculated using the 2 Ϫ⌬⌬Ct method. The TaqMan probes (ThermoFisher Scientific) used in this study are as follows: human BiP (Hs99999174_m1); human ␤-actin (ACTB) (Hs99999903_m1); Drosophila BiP (Dm01813415_g1); and Drosophila GADPH (Dm01841186_g1).

Western blot analysis
Ten adult fly heads or cultured SK-N-MC cell samples (in standard 24-well plates) were homogenized/lysed with 2% SDS buffer with a plastic pestle for protein extraction. Protein concentration was detected using the Pierce TM bicinchoninic acid protein assay kit (ThermoFisher Scientific). Primary antibodies used in Western blot analyses are listed as follows: HA-7 (Sigma; 1:1,000) for the detection of HA-tag; 9B11 (Cell Signaling Technology, Inc., Danvers, MA; 1:1,000) for Myc-tag; and Ab6046 (Abcam, Cambridge, UK; 1:2,000) for

Bisamidine-based inhibitor reduces polyQ aggregation
␤-tubulin. The protein chemiluminescence signal was obtained and visualized with the ChemiDoc TM Touch Gel Imaging System (Bio-Rad). The images were analyzed with ImageLab TM software (Bio-Rad).

Recombinant protein aggregates formation
The recombinant protein aggregate formation in a cellfree condition was previously described (63). In a typical procedure, aggregation was initiated by adding enterokinase (EK) (Ipswich, MA) into 10 M purified recombinant Trx-Htt-Q46 protein to cleavage the Trx tag from polyQ protein.
AQAMAN was added at the time of adding EK (to inhibit formation of aggregates) or after 24 h of adding EK (to dissociate preformed aggregates). The mixture was then incubated at 30°C for 72 h to form the aggregates or to dissociate the preformed aggregates. To stop the reaction, an equal volume of 4% SDS was added to each sample, followed by boiling at 99°C for 10 min. The presence of aggregates was detected using a filter trap assay.

Filter trap assay
SK-N-MC cells cultured in standard 24-well plates were lysed in 2% SDS solution and collected. Protein samples were diluted to a final volume of 200 l with the same solution. The protein sample was then transferred to 48-well Bio-Dot microfiltration apparatus and filtered with cellulose acetate membrane (pore size 0.2 m; Sartorius Stedim Biotech Gmbh, Goettingen, Germany). The membrane was blocked with 5% nonfat dry milk and incubated with antibodies (same antibodies used for Western blottings) for aggregated protein detection.

ITC
ITC-binding assay was performed on MicroCal iTC 200 (Malvern Instruments Ltd., Malvern, UK). AQAMAN was dissolved in autoclaved ddH 2 O and diluted to 6 mM with protein buffer (50 mM Tris, 150 mM NaCl, 1 mM PMSF, 1 mM EDTA, pH 8.0). With the system temperature set at 25°C, 6 mM AQAMAN was titrated against 0.2 mM of Trx protein or Trx-Htt-Q46 protein. Titration results were analyzed with Origin scientific plotting software version 7 (OriginLab Corp.) and fitted with sequential binding sites model.

Statistical analysis
For comparisons between three or more sample groups, analysis of variance with the Tukey post hoc test was performed. For pairwise comparisons, a Student's t test was used. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. NS indicates not significant. All histograms depict mean Ϯ S.E. All experiments were performed at least three times independently.

Institutional animal care and use
The animal research in this study was approved by the Animal Experimentation Ethics Committee of the Chinese University of Hong Kong.