A brain-targeting lipidated peptide for neutralizing RNA-mediated toxicity in Polyglutamine Diseases

Polyglutamine (PolyQ) diseases are progressive neurodegenerative disorders caused by both protein- and RNA-mediated toxicities. We previously showed that a peptidyl inhibitor, P3, which binds directly to expanded CAG RNA can inhibit RNA-induced nucleolar stress and suppress RNA-induced neurotoxicity. Here we report a N-acetylated and C-amidated derivative of P3, P3V8, that showed a more than 20-fold increase in its affinity for expanded CAG RNA. The P3V8 peptide also more potently alleviated expanded RNA-induced cytotoxicity in vitro, and suppressed polyQ neurodegeneration in Drosophila with no observed toxic effects. Further N-palmitoylation of P3V8 (L1P3V8) not only significantly improved its cellular uptake and stability, but also facilitated its systemic exposure and brain uptake in rats via intranasal administration. Our findings demonstrate that concomitant N-acetylation, C-amidation and palmitoylation of P3 significantly improve both its bioactivity and pharmacological profile. L1P3V8 possesses drug/lead-like properties that can be further developed into a lead inhibitor for the treatment of polyQ diseases.

Amino acid substitutions of P3. The observation that favorable entropic change contributes to P3 interaction with expanded CAG RNA suggests that the peptide may have undergone conformational changes upon binding to RNA. Structure prediction of the P3 peptide using the PEP-FOLD server 23 suggested that although P3 does not adopt a well-defined tertiary structure, it may preferentially adopt a loosely folded coil conformation that positions the pharmacophores Lys3, Tyr9, and Phe12 on the same surface of the peptide, which may in turn facilitate the recognition and interaction of RNA ( Supplementary Fig. S1). Such conformation is maintained partially by a network of interaction mediated by the side chain of Asp1. To test whether the orientation of Asp1 is important, we restricted it by substituting Gly2 with D-alanine (P3V1). Our result shows that the binding affinity of P3V1 was improved by 4-fold which gave a K D of 2 µM. Such improvement could be due to better stabilization of the RNA-interacting surface via the restriction of Asp1 or because the replacement of Gly by D-Ala minimized the entropic loss upon binding by restricting the flexibility of the peptide. Analysis of the ITC results reveals that the change in the Gibbs free energy of P3V1 mainly originated from a favorable gain in enthalpy, indicating that the former scenario is more likely. Considering that both Phe12 and Lys13 are important for RNA binding, we kept the 13-mer P3 as our minimal construct for further studies.
To improve the binding affinity and specificity of P3 toward expanded CAG RNA, we performed amino acid substitutions at individual pharmacophores using natural amino acids with similar properties, i.e. Tyr and Phe to other aromatic side chains; Lys to Arg ( Table 1). Substitution of Tyr9 with the larger Trp side chain (P3V2) resulted in a lower K D of 2.2 µM and the substitution of Phe with less hydrophobic Tyr (P3V3) and Trp (P3V4) also improved the binding of peptide to expanded CAG RNA by nearly 2-to 4-fold, respectively. These results indicate that although alterations of the size or polarity of the aromatic residues improved the binding affinity of P3, only modest improvements were obtained. On the other hand, the K D value of P3V5, in which Lys3 and Lys5 both mutated to Arg residues, was approximately 0.86 µM, reflecting a nearly 10-fold improvement over the P3 peptide. Contrarily, when we mutated all three Lys to Arg (P3V6), the K D of the peptide increased to 23 µM, indicating that the replacement of the amine of Lys13 with a guanidinium is not favored.
Termini modifications improve P3 activity. Next, we explored whether modifications at the N-and C-termini could improve the binding and inhibitory activity of P3. N-terminal acetylation and C-terminal amidation neutralize the charges at both ends of the peptide and we speculated that it may increase its biological activity. The C-terminal amide was incorporated by solid-phase synthesis of the peptide on a Rink amide linker, while the N-terminal modifications were introduced after assembly of the peptide and before acidolytic release of the completed structure. The K D value of P3V7 was determined to be 2 µM using ITC, indicating that C-terminal amidation improved the peptide affinity to expanded RNA. When the N-terminus was also acetylated, the dissociation constant of P3V8 was further lowered to 0.33 µM (Table 1). Based on these observations, we further synthesized P3V9, which combined the modifications of P3V8 and P3V5 (N-acylation, C-amidation and Lys-to-Arg mutation) with the hope to further improving the binding affinity of the peptide. To our surprise, the binding affinity of the new peptide was not improved, as the K D value was 2.3 µM (Table 1). We next investigated whether substitution of Lys3 and Lys5 with other non-natural Lys and Arg analogs would improve the activity of P3V8. The homologated Arg analog, homoarginine (hArg), and two shortened Lys analogs, ornithine (Orn) and 2,3-diaminopropionic acid (Dap), were used for substitution in our studies (Supplementary Table S2, P3 variants 14-19 respectively). Our results reveal that when either Lys3 or Lys5 was replaced by any of the amino acid analogs, no binding could be detected between the P3 variant and the expanded CAG RNA using ITC. This indicates that the chain length of Lys in P3V8 is crucial for its bioactivity. In addition, the linear dimer of P3V8, which has two copies of the peptide in close vicinity, did not show any binding (Supplementary Table S2, P3V20 and P3V21). The abolishment of the bioactivity may be due to steric hindrance.
To test whether P3V8 specifically targets expanded CAG RNA, ITC experiments were performed using unexpanded CAG RNA (MJD CAG27 RNA), expanded CAG RNA (MJD CAG78 RNA), and expanded CAA/G RNA (MJD CAA/G78 RNA), respectively, to investigate whether the binding of P3V8 is RNA expansion-dependent and sequence-dependent (Supplementary Table S3). Our results demonstrate that P3V8 bound unexpanded RNA or expanded CAA/G RNA with nearly 6-9-fold lower affinity, which indicates that P3V8 is expanded CAG RNA-specific.
P3 derivatives showed improved suppressive effects on polyglutamine neurodegeneration in vivo. Given that P3V5 and P3V8 bind expanded CAG RNA with significantly higher affinities than P3, we next tested their inhibitory activities against polyQ neurodegeneration in vivo. We previously demonstrated that the full-length MJD CAG transgenic Drosophila model, flMJD CAG27/84 , can be used to investigate the suppression effects of synthetic peptide inhibitors on both expanded RNA and polyQ protein toxicities 21 . Here we utilized the DsRed CAG0/100 Drosophila model to investigate the suppressive effects of the P3 analogs on RNA toxicity. In this model, the CAG repeat is located in the 3′ untranslated region of the DsRed reporter gene 7 . The transcribed expanded CAG RNA is not translated into polyQ protein, and thus the toxicity is solely attributed to the expression of DsRed CAG100 RNA. In brief, the expression of expanded CAG RNA causes severe retinal degeneration in the animal that can be quantified by a pseudopupil assay 7 . Inhibitors that can suppress neurotoxicity will result in the recovery of the number of rhabdomeres within the ommatidium. To test our derivatives, third instar larvae were fed different amounts of peptide dissolved in sucrose solution for 2 hr and then allowed to culture in standard fly food until the flies were 12 days old, at which time they were sacrificed for the pseudopupil assay. We observed that treatment with 50 µM of P3 peptide moderately suppressed neurotoxicity in DsRed CAG100 flies (pseudopupil score, 3.74 ± 0.05; Fig. 1a and b). In contrast, derivatives P3V5 and P3V8 both significantly suppressed neurotoxicity and recovered the number of rhabdomeres with much greater efficiency ( Fig. 1c-f). In particular, P3V8 displayed a high level of potency against the neurotoxicity in DsRed CAG100 flies and achieved a pseudopupil score of 4.32 ± 0.25 in comparison to P3V5's score of 3.64 ± 0.22 after 1 µM treatment ( Fig. 1d and f). Based on the results from the ITC experiments and pseudopupil assay, the peptide derivative P3V8 seems to be the most potent P3 variant that exerts inhibitory activity against expanded CAG RNA-induced neurotoxicity.
P3V8 restored the expression level of pre-rRNA in a fly model expressing expanded CAG RNA. Ribosomal RNA (rRNA) synthesis occurs within the nucleolus 15 . We previously showed that expanded CAG RNA induces nucleolar stress by preventing the nucleolar protein NCL from binding to the upstream control element of the rRNA promoter, leading to the downregulation of rRNA transcription 13 . To investigate whether P3V8 suppressed neurotoxicity by alleviating expanded CAG RNA-induced nucleolar stress in vivo, we measured the expression levels of pre-rRNA in DsRed CAG100 flies treated with the peptide. Real-time PCR analysis showed that P3V8 could restore the transcript level of pre-rRNA in a dose-dependent manner. When DsRed CAG100 flies were treated with 1 μM of P3V8, the pre-rRNA level was fully recovered to the DsRed CAG0 control level (Fig. 2). This confirms that P3V8 subdued polyQ neurodegeneration in vivo by effective mitigation of the expanded CAG RNA-induced nucleolar stress.  P3V8 suppressed expanded CAG RNA-induced cell death by modulating rRNA transcription in vitro. We next investigated whether P3V8 could also mitigate expanded CAG RNA-induced toxicity in mammalian cells. Overexpression of EGFP CAG78 RNA in HEK293 cells has been shown to induce nucleolar stress and apoptosis 21 . To compare the inhibitory activity of P3WT and P3V8, we used the peptide transfection reagent DeliverX (DX) to deliver P3WT and P3V8 to EGFP CAG78 RNA-expressing HEK293 cells. Figure 3a shows that irrespective of the addition of DX transfectant, treatment of P3WT lower than 1 µΜ did not elicit any significant suppression effect on expanded CAG RNA-induced cell death. Although P3V8 alone did not exert any cellular effect, its administration with the aid of DX could effectively alleviate the cytotoxicity induced by expanded CAG RNA (Fig. 3b). To confirm that P3V8 inhibited expanded CAG RNA-mediated nucleolar stress, we measured the levels of pre-45s rRNA in EGFP CAG78 RNA-expressing cells with or without treatment of P3V8 (Fig. 3c). Overexpression of the expanded CAG RNA reduced the level of pre-45s rRNA to approximately 40% of the untransfected control.   Upon P3V8 treatment, a dose-dependent restoration of pre-45s rRNA level was observed (Fig. 3c). Taken together, our results demonstrate that P3V8 could effectively mitigate expanded CAG RNA-induced nucleolar stress and neutralize the cytotoxic effect.
Assessment of P3V8 toxicity. Because we aim to develop P3V8 into a potent inhibitor that targets RNA toxicity in degenerating cells within the nervous system, we assessed whether P3V8 had deleterious effects on neuronal cells by determining lactate dehydrogenase release in the culturing medium of rat cortical neurons. After incubation of P3V8 in concentrations ranging from 0.01 nM to 1 µM for 72 hr, no significant cell death was observed in the P3V8-treated neurons when compared to the untreated control ( Supplementary Fig. S2). It should be noted that the peptide also did not show any observable effect when DsRed CAG0 Drosophila was treated with P3V8 at up to 50 µM ( Fig. 1e and f). This indicates that P3V8 exhibits low cytotoxicity both in vitro and in vivo.
N-terminally acylated P3V8 showed enhanced cellular uptake. Acylation of peptides with long-chain lipids has been shown to improve their cellular uptake, in vivo half-life, and membrane permeability and to change their pharmacokinetic properties [24][25][26] . Importantly, lipid solubility is a key factor for the transportation of inhibitors across the blood-brain barrier. We therefore investigated whether acylation of P3V8 could improve its cellular uptake in vitro as well as bioavailability both in vitro and in vivo. The L1P3V8 peptide was generated by introducing palmitic acid N-terminally to P3V8 in place of N-acetylation (Fig. 4a). LC/MS was used to detect the level of P3V8 or L1P3V8 uptaken by HEK293 cells after 3 hr treatment. Figure 4b shows that higher cellular concentration of L1P3V8 was measured in HEK293 cells when compared to that of P3V8 in both 500 nM and 1 µM treatment groups. While we could detect a low cellular concentration of L1P3V8 in HEK293 cells treated with 100 nM of peptide, no uptake of P3V8 was detected (Fig. 4b). These data collectively demonstrated that the lipidation of P3V8 improved peptide cellular uptake. We next administered L1P3V8 to EGFP CAG78 RNA-expressing HEK293 cells without any DX peptide transfecting reagents to determine whether lipidation can facilitate the cellular uptake of the peptide. As shown in Fig. 4c, L1P3V8 alone effectively inhibited expanded CAG RNA-induced cell death with an empirical IC 50 value of ~100 nM. In addition, L1P3V8 showed no cytotoxic effects on rat cortical neurons and DsRed CAG0 Drosophila model, and no obvious alteration in egg-to-adult viability of wild type flies (Supplementary Figs S3 and S4 and Fig. 4d). Treatment of DsRed CAG100 Drosophila with L1P3V8 also suppressed retinal degeneration ( Fig. 4d and e). These findings indicate that lipidation of P3V8 improved its cell penetration properties without affecting its inhibitory activity.

Lipidation improved P3V8 in vitro stability.
To test whether lipidation could also improve the stability of P3V8, we compared the stabilities of P3V8 and L1P3V8 in rat plasma and brain homogenate via in vitro incubation to estimate their extent of degradation in plasma and brain tissues. The results shown in Table 2 indicate that the stability of L1P3V8 in both rat plasma and rat brain homogenate at 37 °C was significantly improved in comparison to that of P3V8. At the concentration of 2000 ng/mL, only 21% and 3% of P3V8 remained stable in plasma and brain homogenates, respectively, whereas nearly all L1P3V8 remained intact in plasma and around 45% of L1P3V8 could be detected in the brain homogenate after 1 hr of incubation. When the incubation was extended to 3 hr, around 87% and 21% of L1P3V8 were still detectable in the plasma and brain homogenates, respectively, whereas P3V8 was almost completely degraded. Such improved stability of L1P3V8 was also observed in plasma when the experiments were repeated using 1000 ng/mL and 500 ng/mL of L1P3V8. Because the concentrations of both P3V8 and L1P3V8 after incubation at 500 ng/mL with brain homogenates were below the limit of quantitation (LOQ), no conclusion on comparison of their stabilities in brain homogenate could be drawn. Nonetheless, our results demonstrate that lipidation of P3V8 with palmitic acid significantly improved its stability in different biological matrices.
In vivo pharmacokinetic properties and brain uptake of P3V8 and L1P3V8. The plasma concentration versus time profiles of P3V8 and L1P3V8 after intravenous administration at 3 μmol/kg in Sprague Dawley (SD) rats were studied and compared ( Fig. 5a and Table 3). Without lipidation, P3V8 was quickly eliminated in the plasma 10 min after administration and its half-life was too short to be determined. In contrast, L1P3V8 had a significantly longer half-life of 17 min (Table 3) and remained quantifiable at 190 nmol/L in the plasma 90 min after dosing (Fig. 5a). The maximum observed drug concentration (C max ) and integrated time-concentration responses (area under the curve, AUC) of L1P3V8 were also significantly higher than that of P3V8 (Table 3), indicating better plasma stability after lipidation. The ability of P3V8 and L1P3V8 to be taken up by the brain was also analyzed after intravenous bolus injection. It was noticed that no P3V8 and 13 pmol/g L1P3V8 was detected in the rat brain 20 min after administration (Fig. 5b), suggesting that intravenous administration might not be suitable for brain delivery of these peptides. These undesirable pharmacokinetic properties led us to explore an alternative route of delivery of these peptides to the brain. One of the options is intranasal administration. Intranasal delivery involves the externally exposed olfactory or trigeminal nerve systems and thus is the most direct method of noninvasive delivery method to the brain 27 . Both P3V8 and L1P3V8 were administered intranasally into SD rats at a dose of 3 μmol/kg with pretreatment of 0.5% mucoadhesive chitosan. The pharmacokinetic profiles obtained are significantly different from those obtained after intravenous administration ( Fig. 5c and Table 3). It was noted that the plasma levels of the intranasally administered inhibitors were significantly lower than that from intravenous administrations. The concentration of P3V8 in plasma peaked at 15 min and diminished rapidly within 20 min, whereas L1P3V8 concentration peaked at 18 min and remained stay around 9 nmol/mL at 90 min. Similar to the observations from intravenous administration, both C max and AUC of L1P3V8 were significantly higher than those of P3V8 after intranasal administration, which clearly demonstrates again that lipidation of the inhibitor increased its systemic exposure. In addition, significantly higher concentrations of L1P3V8 (58 pmol/g) and P3V8 (15 pmol/g) in the brain were achieved 20 min after intranasal dosing, as compared to those from intravenous administration, indicating better brain uptake via the intranasal route (Fig. 5d). Taken together, our findings support that the combined strategies of lipidation and intranasal administration significantly improve the pharmacokinetic properties and brain uptake of our peptide inhibitor.

Discussion
Despite our growing knowledge of the role of RNA transcripts in the pathological mechanisms of polyQ diseases, there is a lack of inhibitors that can counteract their neurotoxicity. On the basis that sequestration of NCL by expanded CAG RNA induces cytotoxicity, we recently developed a peptide inhibitor P3 that disrupts NCL-RNA  Table 2. Stability of P3V8 and L1P3V8 in different biological matrices after incubation at 37 °C. Data are presented as mean ± S.E.M. for 5 independent experiments. ***Indicates P < 0.001, significant difference compared with P3V8. # Indicates the concentration was below the lowest limit of quantification (12.5 ng/mL for P3V8 and 100 ng/mL for L1P3V8). N/A indicates not applicable. interaction and consequently mitigates RNA toxicity in polyQ diseases. In this study, we determined the equilibrium dissociation constant K D of P3 using ITC and discovered that the binding of P3 to RNA is both enthalpically and entropically favored. Short peptides like the 13-mer P3 usually assume flexible conformations in an aqueous solution and therefore suffer entropic loss upon binding to their targets. The finding that P3 binds expanded RNA with a favorable TΔS suggests that P3 may have undergone a conformation change upon binding to the RNA in solution.
When characterizing the effects of Ala substitutions of the pharmacophores of P3 on its interaction with expanded CAG RNA, we found that the basic Lys residues are more important than the aromatic residues in the interaction with expanded RNA. Substitution of any of the three Lys in P3 greatly reduced its binding affinity to RNA, indicating that electrostatic interactions play a dominant role in the affinity of P3. This is no surprise because CAG-repeat RNA duplex adopts an A' helical conformation that is intermediate to the A-and B-forms of nucleic acids and is highly electronegative on its surface and thus charge-complementary with the key pharmacophores of P3. As described above, the substitution of Lys at positions 3 and 5 by Arg remarkably improved the binding affinity of the peptide by nearly 10-fold (Table 1, P3V5). This observation suggests that the higher pKa of arginine and its guanidinium group may have further strengthened the ionic interaction with the RNA. Surprisingly, when we further mutated Lys13 to arginine, the binding affinity of the peptide was adversely reduced by nearly 3-fold, indicating that the amine group of Lys13, but not the charge alone, is critical for the binding of P3 and RNA. Further research is needed to investigate whether the side chain amine of Lys13 is important for stabilizing the peptide for RNA recognition or for mediating the critical interaction with nucleic acid.
Previous studies have shown that the charge neutralization of the N-and C-termini of synthetic peptides by acetylation and amidation can improve their stabilities in cell cultures and serum, thus improving their biological activity. Based on these findings, we generated P3V8, which is capped by acetylation and amidation at the N-and C-termini, respectively, in an attempt to improve the biological activity of P3 and in hopes that the same strategy can be applied to other P3 derivatives. We observed that capping of both termini of P3 not only improved its binding affinity to expanded CAG RNA (K D =0.33 µM) but also significantly improved its potency both in vivo and in vitro. In a recent study of compounds that block Alzheimer's Aβ channel activity, Flora and colleagues showed that capping of the amine and carboxyl groups of free histidine helps to improve the residue activity by preventing nonspecific interaction with other reactive residues in the target 28 . We speculate that the capping of P3, now termed P3V8, might have exerted a similar effect and prevented nonspecific interaction between the peptide and the RNA's negatively charged surface, thus improving the binding and efficacy of the peptide. Furthermore, the slight increase in pI of the peptide might also have contributed to its better affinity to the electrostatically negative surface of RNA (P3V8, pI = 10.02 vs. P3, pI = 9.64).
With the success of P3V8, we applied the same capping strategy to P3V5 to further improve its activity. However, as described above, the combination of Lys-to-Arg mutations of residues 3 and 5 and the capping of the peptide termini did not improve the peptide's binding affinity but instead reduced the K D to about 2.3 µM. We speculate that upon N-and C-termini modifications, the interactions of the arginines and RNA might have been altered and that P3V9 likely adopts a different binding mode than P3V5 or P3V8. Moreover, the lengths of the basic side chains at residues 3 and 5 also appear to play a determining role in the binding of the peptide, as either lengthening or shortening the basic side chains abolished the interaction between the inhibitors and expanded RNA, suggesting that the distances between the basic side chains and the RNA is critical and that only limited conformational flexibility is allowed after P3V8 binds to the RNA. Such a rigid binding mode of P3V8 is further supported by the observation that tandem repeats of the peptide failed to interact with the expanded RNA. Studies on the conformation and dynamics of P3V8 before and after RNA binding are needed to understand the mechanism of its specific inhibitory activity.
RNAs are becoming more recognized as attractive therapeutic targets because they fold into well-defined secondary and tertiary structures but exhibit a large variety of conformations, which can provide favorable opportunities for specific drug targeting. One interesting feature of the A' form of CAG RNA is the widened major groove caused by the non-canonical AA base pairs 29 . Such widening provides binding sites that are unique to CAG-repeat RNA, and their accessibility may offer opportunity to improve the potency and specificity of our inhibitor. Further structural information on the interaction between P3V8 and CAG RNA will be useful to provide insight into how to make use of such a unique feature. Nevertheless, the results of our study demonstrates that neutralization of the N-and C-termini by simple modifications like acetylation and amidation can have significant effects on the binding and inhibitory properties of peptidyl inhibitor against an RNA target.  In the past few decades, several inhibitors targeting protein toxicity [30][31][32][33] or CAG RNA toxicity 21,34 in polyQ disease have been developed. Some even showed therapeutic potential [30][31][32][33] . In this study, we demonstrated for the first time a structural activity relationship (SAR) investigation of the peptidylic inhibitor, P3, toward CAG RNA toxicity in polyQ diseases. The SAR study led us to identify a more potent peptide derivative of P3, P3V8, which shows dramatically improved inhibitory efficacy against expanded RNA-mediated toxicity in vivo. However, many peptides have short half-life in vivo, typically in minutes, which raises concern whether P3V8 could be administered as a therapeutic agent. Previous studies have established that covalent anchoring of lipids to peptides promotes peptide binding to the fatty acid binding sites on albumin 35 . As albumin has a very long half-life, this 'docking' into albumin can significantly extend the functional half-life of peptides. Furthermore, we speculated that lipidation of a peptide could improve the ability to cross cell membranes and enter cells. We therefore lipidated P3V8 by N-acylation with palmitic acid. Stability studies of the lipidated peptide L1P3V8 in plasma and brain homogenates revealed that lipidation significantly enhanced the cellular uptake and in vivo stability of P3V8. Above all, the lipidation strategy notably improved the pharmacokinetic profile and brain uptake of L1P3V8 in rats when it was administered intranasally, conferring our inhibitor drug/lead-like properties. The ability of therapeutic agents to pass the blood-brain-barrier is one of the most critical requirements for treatment of neuronal diseases. Although the mechanism of brain uptake of L1P3V8 remains to be elucidated, our results illustrate that the combination of lipidation and intranasal administration may provide a new means to improve the brain uptake of therapeutic agents. In conclusion, our results provide proof of concept that lipidated peptidyl inhibitors that target RNA toxicity are a novel therapeutic option for polyQ diseases. Further modification and optimization on this lead-like L1P3V8 will be needed to prolong its half-life and uptake level in the brain.

Methods
Peptide modeling. Molecular modeling of P3V8 was carried out using the online PEP-FOLD 2.0 server 23 .
The sequence of the 13 amino acid peptide was submitted to the server and 100 simulations were performed using the default settings. The program returned the most representation conformations identified in terms of energy and population, and clustered them based on their sOPEP (Optimized Potential for Efficient structure Prediction) coarse grained energies. The representative model of the top-ranked cluster with the lowest sOPEP energy values was selected for this study.
All standard Fmoc amino acids were coupled in DMF with 5.2 equivalents of amino acids and HOAt, with 5 equivalents of HBTU and 9.4 equivalents DIEA. The coupling time was 60 min at room temperature. All non-standard amino acids, as well as palmitic acid, were coupled using 2 equivalents of amino acids and HOAt, 1.9 equivalents of HBTU and 3.6 equivalents DIEA. Coupling times were 10 min at 75 °C. N-terminal fatty acid was introduced using the same conditions as for the coupling of standard amino acids.
In the synthesis of peptides P3V7 to P3V21, the first 10 couplings used a coupling time of 60 min at 25 °C with NMP and washings in-between couplings. Deprotections were performed by treatment with piperidine-DMF (2:3) for 3 min, followed by piperidine-DMF (1:4) for 15 min. After each coupling and deprotection, a washing procedure with NMP (3x), DCM (1x), then NMP (3x) was performed. From the 10th coupling onwards, the coupling time was increased to 2 × 120 min and an extra deprotection step (15 min) with piperidine-DMF (1:4) was added. The N-terminal acetylation was achieved with acetic anhydride in DMF (1:4) for 2 × 15 min. After completion of peptide chain assembly, the resin was washed 6 times with DCM. All the peptides were released and deprotected by treatment with a cocktail of trifluoroacetic acid (TFA), triethylsilane (TES) and H 2 O (95:2:3 or 95:2.5:2.5) for 2 hr. The TFA solutions were concentrated under a flow of nitrogen and the compounds were precipitated with diethylether to yield the crude products. All peptides were purified by RP-HPLC.
The sequences of peptides are listed in Table 1, Supplementary Tables S1 and S2. The purity of peptides used in cell experiments and in vitro binding was over 90%. Desalted peptides were used in Drosophila feeding assays.
SCieNTifiC REPORTS | 7: 12077 | DOI:10.1038/s41598-017-11695-y Isothermal titration calorimetry binding assay. Experiments were carried out using a MicroCal iTC200 isothermal titration calorimeter (GE Healthcare) at 25 °C. Data were analyzed using the Origin ® scientific plotting software version 7 (Microcal Software Inc.). All RNAs and peptides were dissolved in binding buffer (20 mM MOPS, pH 7.0; 300 mM NaCl) and 0.7 mM of peptide was titrated into 0.5 µM of RNA for each experiment. The concentration of RNA was estimated with appropriate extinction coefficients at 260 nm on a Nanodrop 2000 (Thermo Scientific). A reference power of 8 μcal/s was used with an initial 0.5 µl of injection of peptide followed by 2 µl for all subsequent titrations points with a 60 sec initial equilibrium delay and 150 sec pause between injections. The samples were stirred at a speed of 1000 rpm throughout the experiment. The thermal titration data were fitted to the 'one binding site model' to determine the dissociation constant (K D ). Each experiment was repeated at least 3 times with consistent results obtained.
Cell culture, plasmid transfection and peptide transfection. HEK293 cells were cultured at 37 °C with 5% CO 2 in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. Primary rat cortical neurons were isolated and cultured as previously described 37 . Transient transfection of HEK293 cells was performed using Lipofectamine 2000 (Life Technologies). P3V8 were delivered to HEK293 cells using the DeliverX (DX) Peptide Transfection Kit (Affymetrix) 4 hr after DNA transfection. At least two batches of independently synthesized peptides were used in the experiments.
Lactate dehydrogenase (LDH) cytotoxicity assay and IC 50 determination. To detect the effect of P3V8 and lipidated P3V8 (L1P3V8) on inhibiting cell death in EGFP CAG78 RNA-expressing HEK293 cells, a LDH assay was employed. HEK 293 cells were seeded on a 24-well plate at a density of 0.5 × 10 5 , and pEGFP-CAG78 DNA construct was used to transfect the cells. For P3V8 treatment, different concentrations of P3V8 as indicated in the figure were transfected by DeliverX transfectant 4 hr after DNA transfection. For L1P3V8 treatment, different concentrations of L1P3V8 were added into individual wells immediately after DNA transfection. Seventy-two hours after treatment, LDH enzyme activity in the cell culture medium was measured as described before. Experimental groups were normalized to the untransfected control. After normalization, data were analyzed using the dose response-inhibition curve (nonlinear regression-variable slope) to determine the IC 50 value (Prism6 software, GraphPad Software, Inc.). Each experiment was repeated at least 3 times.
Drosophila genetics, peptide feeding and assays. Flies were raised at 21.5 °C on cornmeal medium supplemented with dry yeast. Fly lines bearing UAS-DsRed CAG0 and UAS-DsRed CAG100 were kind gifts of Professor Nancy Bonini (University of Pennsylvania, USA). The gmr-GAL4 fly line was obtained from Bloomington Drosophila Stock Center. For pseudopupil assay, third instar larvae were fed with various amount of respective peptides dissolved in 2% sucrose solution for 2 hr and then continued to culture in standard fly food at 21.5 °C 38 . Pseudopupil assay was performed on 12 day-old adult flies as mentioned previously 39 . Images were captured by SPOT Insight CCD camera controlled by the SPOT Advanced software (Diagnostic instruments Inc.). Image processing was performed using the Adobe Photoshop CS software (Adobe). Each experiment was repeated at least 3 times (n = 10 fly heads), and consistent results were obtained. For viability test of wild-type flies from egg to adult stage, eggs born within 5 hr were collected and cultured in fly food containing 10 or 50 μM of respective peptides at 21.5°C. Viability from egg-to-adult was calculated as the number of adult flies collected divided by the number of eggs examined. Each experiment was repeated for three times (at least 130 eggs were examined in each group). Two batches of independently synthesized peptides were used in the experiments.
RNA extraction, reverse transcription-PCR and real-time PCR. RNA was extracted from cells or ten 12 day-old adult fly heads by Trizol reagent (Life Technologies), and 1 μg of purified RNA was then used for reverse-transcription using the ImPromII ™ Reverse Transcription System (Promega). Random hexamer (Roche) was used as primers in reverse transcription. Taqman gene expression assays were performed on an ABI 7500 Real-time PCR system and data were analyzed as previously described 13 . The following probes were used: pre-45s rRNA (Assay ID: AILJIZM), pre-rRNA (Assay ID: AIMSG5U), Drosophila GAPDH (Assay ID: Dm01841186) and human actin (Assay ID: Hs99999903_m1). Each experiment was repeated at least 3 times.
In vitro stability studies in rat plasma and brain homogenate. Sprague-Dawley (SD) rats (male, 180-200 g) were supplied by the Laboratory Animal Services Centre at The Chinese University of Hong Kong. All animal studies were conducted under the approval of the Animal Ethics Committee of The Chinese University of Hong Kong, and were performed in accordance with relevant guidelines and regulations. Blank rat brain homogenate was prepared by ultrasonic probe homogenization (Micoson XL-2000, Misonix, Framingdale, NY, USA) of brains collected from the control rats. Blank rat plasma was prepared by centrifugation (8000 rcf for 3.5 min) of blood collected from control rats. For the stability test, 2000 ng/mL and 500 ng/mL of P3V8 or L1P3V8 was spiked into the blank rat plasma or brain homogenate and vortexed. The mixtures were incubated at 37 °C with 100 rpm in a water bath for 1 or 3 hr. The incubation was terminated by addition twice the volume of acetonitrile and centrifugation at 13000 rcf for 10 min. The supernatant was collected for LC/MS/MS analysis. Cellular uptake studies in HEK293 cells. HEK293 cells were seeded on a 6-well plate at a density of 6 × 10 5 and cultured overnight at 37 °C with 5% CO 2 in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. Cells were then treated with 100, 500 or 1000 nM of respective peptide at 37 °C for 3 hr. After treatment, the cells were washed with ice-cold PBS and then lysed by 2% SDS solution. Aliquot (100 µl) of the cell lysate was added with 200 µl acetonitrile and centrifuged at 13000 rcf for 10 min to precipitate the proteins. The collected supernatant was subjected to LC/MS analysis. The protein concentration of the cell lysate was measured using a bicinchoninic acid protein assay kit following Sigma's protocol. Plasma pharmacokinetic and brain uptake studies in SD rat. SD rats (180-200 g) were anesthetized with an intraperitoneal injection of ketamine (60 mg/kg) and xylazine (6 mg/kg) and received a minor surgery of cannulation with a polythene tube (0.4 mm i.d., 0.8 mm o.d., Harvard Apparatus, Holliston, MA, USA) in the left jugular vein. The rats received an overnight recovery with free access to food and water. In the following day, rats were administered P3V8 or L1P3V8 (3 μmol/kg) via intravenous or intranasal routes. For intravenous administration, appropriate volume of P3V8 or L1P3V8 (3 μmol/mL, dissolved in water) was injected to the rats via the cannula. For intranasal administration, the rats were temporarily anesthetized by inhalation of carbon dioxide, and was administrated with appropriate volume of P3V8 or L1P3V8 (30 μmol/mL, dissolved in water with 5% PEG400) in both nostrils by a micropipette. Chitosan solution (0.5% w/v, pH 6.5, 20 μL) was pre-treated to rats intranasally 5 min before the P3V8 and L1P3V8 was administrated via intranasal routes. For the pharmacokinetic study, after the drug administration, blood samples were collected from the catheter at appropriate time intervals (2,5,10,15,20,30,45, 60, and 90 min). After each collection, 0.2 mL of saline containing 25 IU heparin was injected to compensate for the blood loss. Plasma was collected after centrifugation at 8000 rcf for 3.5 min and stored at −80°C until analysis. For brain uptake study, at 20 min, the rat was anesthetized and perfused by 500 mL saline, and the whole brain was collected. The brain was wiped by tissue paper to remove excess water, meninges and blood vessels followed by storage at −80 °C until analysis.
The mixture was vortexed for 1 min and followed by centrifugation at 13000 rcf for 10 min. The supernatant was collect and subjected to LC/MS/MS analysis.
To prepare brain homogenate, the whole brain of each rat was minced. Saline (2 mL/g brain) was added followed by ultrasonic probe homogenization on ice. For analysis of P3V8, aliquot of brain homogenate (100 μL) was mixed with 200 μL acetonitrile. The mixture was vortexed for 1 min and followed by centrifugation at 13000 rcf for 10 min to collect the supernatant. For analysis of L1P3V8, aliquot of brain homogenate (650 μL) was centrifuged at 6000 rcf for 10 min. After the supernatant was collected, 400 μL saline was added to re-suspend the pellet. The mixture was again vortexed for 5 min and centrifuged at 6000 rcf for 10 min. The supernatant collected was loaded to a preconditioned Oasis ® HLB cartridges (Waters, MA, USA). After washed with 1 mL of 10% acetonitrile in water, the analyte was eluted with 0.25 mL 95% acetonitrile in water.

LC/MS/MS analysis. Agilent 6430 Triple Quadrupole LC/MS/MS system (Agilent Technologies, CA, USA)
was employed for the analysis. For analysis of P3V8, chromatographic separation was achieved on a SunFire C8 Column (250 mm × 4.6 mm, 5 µm). The mobile phase of water containing 0.2% formic acid (A) and acetonitrile (B) was used with a gradient elution (0-5 min, 20-70% B). The flow rate was 0.8 mL/min. Multiple reactions monitoring (MRM) with fragmentation transition of 500 to 129 in positive ion mode was employed for quantization of P3V8. For analysis of L1P3V8, Alltima Amino Alltech Column (250 mm × 4.6 mm, 5 µm) was used for separation. Mobile phase of water (A) and acetonitrile (B) was used. The gradient elution for plasma samples was 0-4 min, 30-95% B, and for brain samples was 0-9 min, 10-95% B. The flow rate was 0.8 mL/min. MRM transition of 565 to 88 in positive ion mode was used for quantization of L1P3V8. Data analyses. Plasma concentration verse time profiles were analyzed by WinNonlin (Pharsight Corporation, Mountain View, CA, USA, Version 2.1) to obtain the pharmacokinetics parameters. Other data were analyzed by one-way ANOVA followed by post hoc Tukey test or unpaired t-test. "*", "**", "***" and "****" represent P < 0.05, P < 0.01, P < 0.001 and P < 0.0001 respectively, which are considered statistically significant.