Allosteric Modulation of Pathological Ataxin‐3 Aggregation: A Path to Spinocerebellar Ataxia Type‐3 Therapies

TL;DR

A new therapy for spinocerebellar ataxia type 3 is proposed by targeting a specific protein site to reduce harmful protein clumping.

Contribution

The discovery of a novel allosteric site in ataxin-3 for therapeutic targeting using CLR01.

Findings

CLR01 binds to a lysine residue in the Josephin domain of Atx3, reducing aggregation-prone conformational fluctuations.

CLR01 delays amyloid fibril formation and reduces secondary nucleation rates in Atx3.

CLR01 improves synaptic function in neurons and delays disease progression in animal models.

Abstract

Spinocerebellar ataxia type 3 (SCA3) is a rare neurodegenerative disorder caused by the expansion of a polyglutamine (polyQ) repeat in ataxin‐3 (Atx3) for which no disease‐modifying therapies are available. The presence of protein inclusions enriched in polyQ‐expanded Atx3 in neurons suggests that inhibiting its self‐assembly may provide targeted therapies. Here, it is demonstrated that the supramolecular tweezer CLR01 binds to a lysine residue on a positively charged patch of the Atx3 catalytic Josephin domain, decreasing conformational fluctuations of the distal helical hairpin, without altering its ubiquitin hydrolase activity. This reduces exposure of the aggregation‐prone region that initiates Atx3 self‐assembly, ultimately delaying Atx3 amyloid fibril formation and reducing the secondary nucleation rate, a process linked to fibril proliferation and toxicity. CLR01's effects translate into the reversal of synapse loss in SCA3 cultured cortical neuron model, improve locomotor function in a Caenorhabditis elegans SCA3 model, and delay disease onset with reduced severity of motor symptoms in a SCA3 mouse model. These insights reveal a novel allosteric site for developing CLR01‐inspired therapies targeting pathological aggregation while preserving essential functional sites. They also highlight that targeting allosteric sites in amyloid‐forming proteins may provide new opportunities for safe therapeutic strategies for various protein misfolding disorders.

This study uncovers a new allosteric site in the Josephin domain of ataxin‐3 targeted by the molecular tweezer CLR01, which modulates protein aggregation, improves synaptic function in neuronal cells, and delays motor dysfunction in animal models. This highlights allosteric modulation as a promising strategy for spinocerebellar ataxia type 3 and related protein‐misfolding diseases. Image created in BioRender (https://BioRender.com/oepqx3w).

Figures

CLR01 interacts with Atx3 and induces conformational changes in the JD. a) Chemical structures of the molecular tweezer CLR01 and the control molecule CLR03. b) ESI‐MS spectra of wild‐type Atx3 (Atx3 13Q) and CLR01 at a 1:1 molar ratio. The insets shows the deconvoluted spectra and highlight the two CLR01 binding events. c) Cartoon representation of three states of JD: experimental NMR models of the JD in open (PDB ID: 1YZB,[ 46 ] yellow) and closed (PDB ID: 2AGA,[ 93 ] orange) conformations. A representative structure (light pink) of the “half‐open” conformation of the JD is shown (most populated cluster from the GaMD simulations in complex with CLR01). d) Amino acid sequence of JD with residues undergoing chemical shifts in the NMR HSQC spectra upon CLR01 binding in red. Secondary structure elements in the JD are shown below the amino acid sequence with helix α1 in green, the aggregation‐prone segment (α4–β1) in bold and orange, and helix α6 in teal. e) Cartoon representation of the JD open conformation with the residues affected by chemical‐shift‐perturbation (CSP) upon addition of CLR01 shown as red spheres. K128, a key residue in the interaction between CLR01 and JD, is highlighted as pink spheres. Other regions are colored as in panel (d). f) Cartoon representation of the third most populated cluster from the GaMD simulations of JD (open conformation) with a single CLR01 molecule (magenta sticks) bound at K128 (pink spheres). Residues displaying a chemical shift in the HSQC spectrum are highlighted as red spheres; R182 and other positively charged residues, corresponding to predicted CLR01 binding sites according to GaMD simulations and Gibbs binding energy estimations, are shown as sticks. This structure highlights a state where the hairpin (yellow surface) moves toward the aggregation‐prone region (orange).

CLR01 binding modulates the Atx3 aggregation pathway, delaying amyloid fibril assembly and decreasing the secondary nucleation rate. a,d) Representative ThT progress curves of 5 µm Atx3 13Q (a) or Atx3 77Q (d) amyloid self‐assembly kinetics in the absence or presence of CLR01 or CLR03. Error bars correspond to the standard deviation of three replicates. b,c) Bar plots show mean ± standard error of the mean (SEM) for (b) Atx3 13Q amyloid formation halftime (t 50) and (c) aggregation rate (v 50) derived from fitting of ThT fluorescence curves.[ 99 ] Control samples were compared with CLR01 or CLR03‐treated samples at the indicated molar ratios. Individual points represent independent replicates (N = 5; each representing the mean values from 3–5 technical replicates). e,f) Same as (b, c) for Atx3 77Q, with 4 independent replicates. Statistical analysis was performed using a linear mixed‐effects model (LME) for Atx3 13Q (b, c), and repeated‐measures one‐way ANOVA for Atx3 77Q (e,f), followed by Dunnett's multiple comparisons test (Table S2, Supporting information). g) Time‐dependent DLS intensity distributions during the aggregation of 5 µm Atx3 13Q alone or in the presence of 25 µm CLR01 or CLR03. Different colors show different time points (size distributions measured in triplicate). Vertical lines: scattering intensities predicted by a nucleation‐and‐growth model for monomers (thinner lines) and amyloid fibrils (thicker lines) at the end of 48 h incubation (refer to Figure S7 in the Supporting Information, for other time points); h) same as (g) for Atx3 77Q. i) Fitted nucleation‐and‐growth rate constants for Ax3 13Q and Atx3 77Q; error bars correspond to standard errors (Table 1).

CLR01 reduces the assembly of mature fibrils of mutant Atx3 and dissociates preformed protofibrils and mature fibrils. a) Morphological analysis of 5 µm Atx3 13Q or Atx3 77Q aggregates in the absence or presence of 50 µm CLR01 or CLR03 at the endpoint of the aggregation assay (66 h), monitored by negative staining TEM. Scale bars correspond to 100 nm; orange rectangles denote Atx3 curvilinear protofibrils; red rectangles highlight small protofibrils formed in the presence of CLR01, red triangles point to Atx3 spherical oligomers; white arrowheads mark clusters of spherical aggregates; and white open arrows point to mature Atx3 Q77 fibrils. b) Particle size histograms and c) scatter plots of all particle measurements with median ± 95% CI, from the TEM images shown in (a) (3–4 images per condition were analyzed). Descriptive statistics are shown in Table S3 (Supporting information). d) Morphological analysis of 5 µm Atx3 13Q or Atx3 77Q protofibrils formed at the endpoint of the aggregation assay (t = 66 h), before (T0) and after three days of incubation (T3d) at 37 °C in the absence (Control) or presence of 400 µm CLR01 or CLR03 by negative‐staining TEM. Scale bars correspond to 100 nm. e) Morphological analysis of 5 µm 77Q mature SDS‐resistant fibrils from a 160 h aggregation assay before (T0) and after ten days of incubation (T10d) at 37 °C in the absence (Control) and presence of 400 µm CLR01 by negative‐staining TEM. Scale bars correspond to 200 nm.

CLR01 reverts the loss of glutamatergic synapses in cortical neurons caused by pathological Atx3. Rat cortical neurons were transfected with plasmids encoding eGFP, wild‐type eGFP‐Atx3 28Q, or mutant eGFP‐Atx3 84Q. a) Neurons were immunolabeled for MAP2, PSD95, and vGLUT1. Excitatory synapses were detected as PSD95 puncta that colocalized with vGLUT1; scale bar = 10 µm. b) Integrated fluorescence intensity of PSD95 puncta colocalized with vGLUT1 (n = 33–36 neurons per condition in 3 independent experiments). Statistical analyses were performed using Kruskal–Wallis and Dunn's post‐hoc test (Table S4, Supporting Information). c) Neurons incubated with 10 µm CLR03 or CLR01 were immunolabeled for MAP2, PSD95, and VGluT1 clusters; scale bar = 10 µm. d) Integrated fluorescence intensity of PSD95 puncta colocalized with VGluT1 (n = 35–41 neurons per condition in each of 3 independent experiments. Statistical analyses were performed using Kruskal–Wallis and Dunn's post‐hoc tests (Table S5, Supporting Information). In panels (c) and (d), boxes show the 25th and 75th percentiles, whiskers range from the minimum to the maximum values, and the horizontal line shows the median value. Mean is represented by “+.” e) GFP‐Atx3 accumulates in the cell body of a fraction of transfected cortical cultures treated with CLR03 or CLR01; scale bars = 300, 15 µm. f) Percentage of cells with GFP‐Atx3 aggregates (n = 3 independent experiments, 31–35 neurons per condition). Statistical analyses were performed using two‐way ANOVA and Sidak's multiple comparisons post‐hoc tests (Table S6, Supporting Information). p‐values are indicated in the graphs shown in panels (b), (d), and (f). Data are represented as mean ± SEM.

CLR01 treatment alleviates motor deficits in a C. elegans model of SCA3 even when treatment is initiated at a postsymptomatic disease phase. a) CLR01 treatment improved motor dysfunction of mutant Atx3 animals. AT3q130 animals were treated with different concentrations of CLR01 for four days, and the percentage of locomotion‐defective animals was determined. CLR01 significantly impacted motor phenotype between 0.0001 and 1 µm; the maximum efficiency (%) was observed at 0.1 µm at which 67% phenotypic rescue was achieved compared to the control strain (Table S8, Supporting Information). b) Treatment with 0.1 µm CLR01 for four days did not impact motor phenotype in the wild‐type strain (N2). c) Postsymptomatic treatment with CLR01 improves the animals' motor defects after 2 or 4 days of treatment (days 6 and 8 after hatching). (a–c) Data are represented as mean ± SD of 4 independent experiments, with ≈50 animals per condition/per assay (total number of animals of ≈250). A one‐way ANOVA test was applied, followed by a bilateral Dunnet test for post‐hoc comparisons (a and c). In the comparison between the two groups (b and c before treatment), independent‐samples t‐test was used (Tables S9, S10, Supporting Information). d–g) Impact of early life chronic CLR01 treatment on mutant human Atx3 aggregation pattern, labeled with an Atx3‐specific antibody in the C. elegans model of SCA3 throughout animals’ development and adulthood. Pictures of the animal's heads were obtained by confocal microscopy (Olympus FV1000) and analyzed using the MeVisLab software.[ 106 ] The graphs show a pool of 3 independent assays, with at least four animals per condition in each trial. p‐values were calculated using independent‐samples t‐test (Table S10, Supporting Information). Scale bars = 20 µm.

CLR01 chronic administration improves motor function and neuropathology of SCA3 mice. a) The latency to cross a swimming tank, representing both the strength and the movement coordination ability of the animals, is higher in vehicle‐treated SCA3 mice. The severity of this symptom improves over time with CLR01 treatment, which has a delayed onset and is effective up to 16 weeks of age. b) CLR01 improves SCA3 mice balance, decreasing their latency to cross a 12 mm squared beam. c) The gait quality of SCA3 animals is strikingly improved by CLR01 treatment, delaying the onset of ataxic gait by eight weeks. d) Representative images of cresyl violet staining in the cervical spinal cord of healthy and pyknotic cells. e) Vehicle‐treated SCA3 mice showed a substantial increase in the number of pyknotic cells, which was rescued by daily CLR01 administration for 18 weeks. f) Representative pictures of the motor‐neuron cell bodies (black arrows) in the ventral horn of the cervical spinal cord. g) CLR01 treatment increases the number of motor neurons in the cervical spinal cord. Sample size and statistical details are provided in Table S11 (Supporting Information). The scale bar for lower magnification pictures represents 200 µm, and for higher magnification, 100 µm. p‐values are shown in the graph. (a, b) Data are represented as mean ± SEM (n = 13–19 per group). A repeated measures ANOVA test was applied, followed by Tukey test for multiple comparisons. (c) Data are represented as the % of animals within each category (n = 13–19 per group). A Kruskal–Wallis H test was applied. (e and g) Data are represented as the median (min; max) with all datapoints shown (N = 5–6 animals per group; n = 4–7 slices per mouse). A one‐way ANOVA test was applied, followed by a Tukey test for multiple comparisons. Black lettering represents the comparisons between WT‐vehicle and SCA3‐vehicle. Red lettering represents the comparisons between SCA3‐vehicle and SCA3‐CLR01.

Schematic representation of the effects of CLR01 on Atx3 self‐association. Binding of CLR01 to K128 (colored dark blue) within a positively charged surface patch opposite to the aggregation‐prone region and ubiquitin binding site (orange) triggers a conformational change in the helical hairpin (yellow) of the Josephin domain. The CLR01‐induced conformational change reshapes the Atx3 aggregation pathway and traps spherical oligomers while reducing secondary nucleation rates. This modulation of the self‐assembly pathway correlates with the reversion of synapse dysfunction in SCA3 cell models and a delay in the onset of motor impairment in SCA3 animal models. Image created in BioRender (https://BioRender.com/lgwm68m).

Tables

Table 1. Rate constants estimated from the analysis of the DLS results. Standard errors of fitted parameters are indicated in parentheses.

DLS results	R 1 [nm]	R 2 [nm]	kn Δµ0 × 105 [h−1]	k +Δµ0 × 102 [h−1]	k 2Δµ0 [h−1]	R 1 [nm]	R 2 [nm]	kn Δµ0 [h−1]	k +Δµ0 × 102 [h−1]	k 2Δµ0 [h−1]
Control	150	15	12.3 (42.2)	0	0.14 (0.14)	150	20	1.9  ×  10−4 (3.3  ×  10−3)	0	0.13 (0.79)
CLR01 (5:1)	150	10	0.25 (0.20)	8.7 (2.7))	3.0  ×  10−6 (1.4  ×  10−6)	150	20	2.6  ×  10−6 (1.8  ×  10−6)	9.49 (0.85)	5.6  ×  10−6 (2.0  ×  10−6)
CLR03 (5:1)	100	15	100 (548)	0	0.058 (0.17)	40	20	0.10 (7.12)	0	6.1  ×  10−4 (4.6 ×  10−2)

Full text

Introduction

1

Spinocerebellar Ataxia type 3 (SCA3), also known as Machado–Joseph disease (MJD), is a rare, neurodegenerative, hereditary ataxia with no disease‐modifying treatments. SCA3 results from the expansion of a trinucleotide CAG repeat in the ATXN3 gene, which is translated into an expanded polyglutamine (polyQ) tract in the ataxin‐3 protein (Atx3). Atx3 is a modular deubiquitinating enzyme displaying high conformational plasticity, which is composed of a globular catalytic Josephin domain (JD), a flexible C‐terminal tail containing the polyQ tract, and two or three ubiquitin‐interacting motifs.^[^ 1 ^]^ Aggregates containing Atx3 with pathological polyQ tract lengths (>61Q) represent a key disease fingerprint and accumulate in degenerating brain regions, including the cerebellum, brain stem, substantia nigra, and striatum, as well as the spinal cord.^[^ 2, 3, 4 ^]^ The exact cause of neurotoxicity is still debated, but several cell and animal disease models show that abnormal aggregation of pathological Atx3 contributes to the underlying neuronal damage in SCA3^[^ 2, 5, 6, 7, 8, 9 ^]^ and other polyQ disorders.^[^ 9, 10, 11, 12, 13, 14, 15 ^]^ As a result, tackling aberrant Atx3 self‐assembly is one of the main approaches pursued to find disease‐modifying therapies for SCA3 and other polyQ diseases.^[^ 16, 17, 18, 19 ^]^

Biophysical studies have shown that both wild‐type and pathological forms of Atx3 can assemble into amyloid‐like structures in vitro.^[^ 20, 21, 22 ^]^ An aggregation‐prone region located in the JD (^73^GFFSIQVISNALKVWGLELILFNS^96^), overlapping with the substrate ubiquitin binding site, plays an important role in initiating the Atx3 self‐assembly process^[^ 20, 21, 22 ^]^ and is found in the core of protofibrils formed by both the wild‐type and polyQ‐expanded protein.^[^ 22 ^]^ However, only the expansion of the polyQ beyond the pathological threshold triggers the formation of mature, sodium‐dodecyl‐sulfate (SDS)‐resistant fibrils.^[^ 23, 24, 25 ^]^ Studies using nonpathological Atx3 demonstrated the occurrence of transient populations of soluble oligomers in parallel with the thermodynamically favored amyloid pathway.^[^ 26, 27 ^]^ Given the complexity of the Atx3 self‐assembly process and the heterogeneity of its assembly states, which include monomers, dimers, amyloidogenic and nonamyloidogenic oligomers, amyloid protofibrils, and bundled amyloid fibrils, targeting Atx3 self‐assembly is challenging. Therefore, to develop effective inhibitors that target neurotoxic Atx3 aggregation, dissecting their molecular mechanism of action and comprehending how they interfere with the various steps involved in Atx3 self‐assembly is crucial.

Here, we investigated the molecular tweezer CLR01 (Figure 1a), a Janus‐type supramolecular ligand that possesses an electron‐rich cavity and adjacent phosphate anions that can accommodate a lysine or an arginine side chain, and a convex exterior capable of docking onto hydrophobic clefts on the target proteins.^[^ 28, 29, 30 ^]^ Given its unique recognition mode, CLR01 has been shown to act as a broad‐spectrum inhibitor of abnormal protein self‐assembly.^[^ 31, 32 ^]^ It effectively inhibits the aggregation of various amyloid‐forming proteins linked to neurodegenerative diseases, including amyloid β‐protein (Aβ),^[^ 32, 33 ^]^ tau,^[^ 32, 34 ^]^ α‐synuclein,^[^ 35, 36 ^]^ and the polyQ‐expanded huntingtin exon‐1 protein.^[^ 37 ^]^ The efficacy of CLR01 in reverting disease‐related symptoms in animal models has also been demonstrated for Aβ,^[^ 38, 39 ^]^ tau,^[^ 40 ^]^ and α‐synuclein.^[^ 36, 41 ^]^ By combining in silico, in vitro, and in vivo studies, we evaluated the impact of CLR01 on Atx3 self‐assembly and investigated its effect on disease‐related phenotypes in cell and animal models of SCA3. Our results show that CLR01 binds to a surface patch close to the termini of the JD domain of Atx3, acting as an allosteric modulator that restrains the conformational fluctuations required to expose the distal aggregation‐prone segment, which is part of Atx3 amyloid fibril core.^[^ 25 ^]^ CLR01 effectively remodeled the Atx3 aggregation pathway and not only reduced the rate of secondary nucleation, a crucial step for the proliferation of amyloid fibrils, but also disrupted preformed fibrils. In addition, CLR01 reverted synaptic defects in primary neurons expressing polyQ‐expanded Atx3 and reduced motor deficits in a SCA3 Caenorhabditis elegans model. To further validate its efficacy against Atx3, we tested CLR01 using the CMVMJD135 transgenic mouse model of SCA3, which exhibits quantifiable motor symptoms similar to those observed in patients with SCA3.^[^ 6 ^]^ Chronic and early symptomatic treatment with CLR01 demonstrated significant therapeutic benefits, delaying symptom onset and improving motor function in correlation with rescued motor‐neuron pathology. These outcomes underscore the potential of this newly identified CLR01 binding site as a valuable target in the development of innovative lead molecules to address the toxicity linked to pathological Atx3 aggregation in SCA3.

*CLR01 interacts with Atx3 and induces conformational changes in the JD. a) Chemical structures of the molecular tweezer CLR01 and the control molecule CLR03. b) ESI‐MS spectra of wild‐type Atx3 (Atx3 13Q) and CLR01 at a 1:1 molar ratio. The insets shows the deconvoluted spectra and highlight the two CLR01 binding events. c) Cartoon representation of three states of JD: experimental NMR models of the JD in open (PDB ID: 1YZB,[

46

] yellow) and closed (PDB ID: 2AGA,[

93

] orange) conformations. A representative structure (light pink) of the “half‐open” conformation of the JD is shown (most populated cluster from the GaMD simulations in complex with CLR01). d) Amino acid sequence of JD with residues undergoing chemical shifts in the NMR HSQC spectra upon CLR01 binding in red. Secondary structure elements in the JD are shown below the amino acid sequence with helix α1 in green, the aggregation‐prone segment (α4–β1) in bold and orange, and helix α6 in teal. e) Cartoon representation of the JD open conformation with the residues affected by chemical‐shift‐perturbation (CSP) upon addition of CLR01 shown as red spheres. K128, a key residue in the interaction between CLR01 and JD, is highlighted as pink spheres. Other regions are colored as in panel (d). f) Cartoon representation of the third most populated cluster from the GaMD simulations of JD (open conformation) with a single CLR01 molecule (magenta sticks) bound at K128 (pink spheres). Residues displaying a chemical shift in the HSQC spectrum are highlighted as red spheres; R182 and other positively charged residues, corresponding to predicted CLR01 binding sites according to GaMD simulations and Gibbs binding energy estimations, are shown as sticks. This structure highlights a state where the hairpin (yellow surface) moves toward the aggregation‐prone region (orange).*

Results

2

CLR01 Binds to Atx3, Inducing Conformational Changes in the Helical Hairpin

2.1

Several studies demonstrated that JD dimerization primes Atx3 self‐assembly^[^ 20, 22, 25 ^]^ and that fibril formation is critically dependent on the exposure of an amyloidogenic region that includes an exposed lysine residue.^[^ 21, 24, 25 ^]^ As CLR01 (Figure 1a) can bind to exposed lysine or arginine residues, we hypothesized it may disturb the electrostatic and hydrophobic interactions within the aggregation‐prone region, thereby interfering with Atx3 aggregation.

To evaluate CLR01's interaction(s) with Atx3 and gain insights into its impact on Atx3 conformational dynamics, we conducted native electrospray ionization mass spectrometry (ESI‐MS) studies, alongside Gaussian accelerated molecular dynamics (GaMD) simulations.^[^ 42 ^]^ The addition of CLR01 at a 1:1 ratio to wild‐type Atx3 (Atx3 13Q) revealed the formation of Atx3:CLR01 complexes with a K D = 8 µm, which appeared as distinct distributions with shifts to higher m/z values for each charge state (Figure 1b). Upon deconvolution of the mass spectra, two binding events were observed, reflecting the formation of 1:1 and 1:2 Atx3:CLR01 complexes. Additionally, we observed that similar complexes were formed when CLR01 was added to the isolated JD (K D = 12 µm) or to a disease‐related variant in which the polyQ tract was expanded to 77Q (Atx3 77Q, K D = 14 µm) (Figure S1a,b, Supporting Information). The formation of CLR01‐bound complexes with the isolated JD indicates that CLR01 molecules bind to this domain despite several potentially highly exposed K/R residues in the Atx3 C‐terminal tail. Ion mobility coupled to mass spectrometry experiments indicated that wild‐type and pathological Atx3, as well as the isolated JD, bound to CLR01, shifted toward shorter drift times, i.e., formed more compact conformers (Figure S1c,d, Supporting Information), as observed previously for Aβ42^[^ 33 ^]^ and α‐synuclein.^[^ 43 ^]^

To further understand the impact of CLR01 binding on Atx3 conformational fluctuations, we performed GaMD^[^ 42 ^]^ simulations. Without CLR01, JD (PDB ID: 1YZB) had an open conformation, with the helical α2–α3 hairpin distant from the catalytic subdomain.^[^ 44, 45 ^]^ CLR01 molecules were placed around the surface accessible lysine and arginine residues, resulting in a 1:9 CLR01 complex (Figure S2, Supporting Information). In the presence of CLR01, the system evolved to a “half‐open” conformation, representing an intermediate state between the experimentally reported “open” (PDB ID: 1YZB) and “closed” (PDB ID: 2AGA) conformations (Figure 1c; Figure S3a–d, Supporting Information). Therefore, we also performed analogous GaMD simulations with the closed state (PDB ID: 2AGA) as the starting geometry. Based on our simulations, it is evident that in the absence of CLR01, the JD is better represented by the open conformation (PDB ID: 1YZB), in agreement with previous studies.^[^ 44, 46 ^]^ Additionally, the JD in the closed conformation also transitions to a “half‐open” configuration in the presence of CLR01 (Figure S3d, Supporting Information). As a result, the open conformation was selected for further computational studies. The analysis of the root‐mean‐square‐fluctuation (RMSF) for each residue over the complete GaMD trajectories confirmed that the largest conformational variability is located on the hairpin region of the JD (Figure S3a, Supporting Information).

Taken together, we found that CLR01 binds to JD and modulates the distribution of Atx3 conformers toward intermediate structures between the open and closed states, likely mediated by changes in the mobility of the JD helical hairpin.

The Primary Site of CLR01 Binding on JD Is K128

2.2

To finely map the CLR01 binding sites on Atx3, we combined nuclear magnetic resonance (NMR) analysis with GaMD simulations and Gibbs energy estimations. We specifically focused on the effects of CLR01 on JD because this domain is known for driving the early stages of Atx3 aggregation^[^ 22, 24, 25, 47 ^]^ and binds CLR01 (Figure 1b; Figure S1a, Supporting Information). Enhanced sampling simulations of the JD in complex with nine CLR01 molecules bound to the R/K residues that were sterically accessible (Figures S2, S4a, Supporting Information) suggested that, as a general trend, the lysine residues form more conserved complexes compared to arginine residues (Figure S4b, Supporting Information). Further, we estimated binding Gibbs energy changes for the complexes between CLR01 and selected residues of JD using the central limit free energy perturbation (CL‐FEP approach).^[^ 48 ^]^ The CL‐FEP approach allows evaluating the FEP identity directly from the energy samples of the end states of the system transformation obtained from MD simulations in explicit solvent. CL‐FEP has been shown to deliver excellent accuracy in estimating the binding Gibbs energy of molecular and biomolecular complexes.^[^ 48 ^]^ Among others, CL‐FEP has found applications for studying the effect of supramolecular tweezers on the formation of Staphylococcus aureus biofilm^[^ 49 ^]^ and on the folding stability of a superoxide dismutase.^[^ 50 ^]^ The Gibbs energy calculations reveal seven sites where binding of CLR01 to exposed lysine and arginine residues is favored (Figure S4c, Supporting Information). NMR titration of JD with CLR01 showed that the tweezer induced several clear chemical‐shift‐perturbation (CSP) effects on specific resonances. The residues most affected, indicated by the total disappearance of the amide proton resonance, are E2, G127, K128, I143, and Q179. The resonances of S3, I4, S12, H38, A49, D105, N108, K125, W130, E141, E173–Q176, I181 are perturbed and shifted (Figure 1d; Figure S4d, Supporting Information). The main binding site is centered on the positively charged residue K128, which is located in the β2–β3 loop, next to helix α7 (^174^ADQ^176^), the N‐terminal residues ^2^ESI^4^, and the C‐terminal portion of strand β5 (I143). Although distant in the sequence, most of the residues are clustered in space in the same region, which contains the N‐ and C‐termini of the JD. The CLR01 interaction site includes an exposed hydrophobic patch formed by the aliphatic moieties of Q176 and L177 side chains, which likely establish additional hydrophobic interactions with K128‐bound CLR01. Binding to CLR01 is in agreement with the estimated Gibbs energy changes, showing that the binding of CLR01 to the exposed K128 residue on the JD was thermodynamically favored (Figure S4c, Supporting Information). Residues, G127, W130, I143, and A174, identified as chemically shifted amino acids by NMR analysis (Figure 1d,e), displayed some of the lowest scores of solvent‐accessible surface area (SASA) (Figure S4a, Supporting Information), which suggests that they are buried in the protein and may be indirectly affected by the binding of CLR01 to K128, given that they are in the vicinity of that binding site (Figure 1e). Analysis of SASA, hydrogen bond interactions, and electrostatic potential surface in simulations of JD support K128 as the preferred CLR01 binding site (Figure S4a,f,g, Supporting Information). Indeed, the surface charge of JD is largely asymmetric, and the region surrounding K128 is embedded in a surface patch rich in positive charges, which likely justifies the initial recruitment of CLR01 driven by favorable electrostatic interactions (Figure S4g, Supporting Information). Once K128 is encapsulated, CLR01 initial binding is further stabilized by additional specific interactions with nearby residues. Despite being thermodynamically favored, the binding to other surface‐exposed lysine residues, including K85 in the aggregation‐prone region, may be impeded by the formation of salt bridges and hydrogen bond interactions with neighboring residues (Figure S4f, Supporting Information). GaMD simulations with CLR01 bound exclusively at position K128 showed a reduction of the flexibility of the hairpin compared to JD alone, as shown in the RMSD plot and B‐factor representations of RMSF (Figure S5a,b, Supporting Information). This reduction in flexibility, along with the movement of the hairpin toward the aggregation‐prone region, leads to the formation of more compact structures, as evidenced by the clustering analysis (Figure S5c, Supporting Information), and in agreement with the ESI‐MS data (Figure S1c–e, Supporting Information). Accordingly, changes in the SASA values were found for both regions, consistent with the fact that the hairpin and aggregation‐prone regions are near each other, at least during part of the simulations (Figure S5d, Supporting Information). A slight reduction in residue fluctuation was also observed in α5 (Figure S5b, Supporting Information). This reduced fluctuation may enhance CLR01 binding to alternative positively charged residues in this region, which could explain the 1:2 Atx3:CLR01 complexes observed by ESI‐MS (Figure 1b). We further demonstrate that CLR01 does not interfere with Atx3 ubiquitin hydrolase activity (Figure S5e, Supporting Information), thereby providing additional support for the identification of the CLR01 interaction site, which is distant from the ubiquitin substrate binding site.

Our data show that CLR01 preferentially interacts with K128 at the ^124^RKLGKQW^130^ site, spatially close to the two JD termini. We propose that this interaction could disrupt early self‐assembly events in the JD through allosteric regulation of the flexibility of the helical hairpin that results in the reduced exposure of amyloidogenic helix α4.^[^ 24, 25 ^]^

CLR01 Delays Atx3 Self‐Assembly and Interferes with Secondary Pathways in Atx3 Amyloid Fibril Formation

2.3

Numerous studies have delved into the molecular details of Atx3 aggregation mechanism, and the influence of the flexible tail, which contains the pathological polyQ expansion, on the local dynamics of the aggregation‐prone JD.^[^ 23, 24, 25 ^]^ Given the ability of CLR01 to interact with JD modulating the flexibility of regions critical for both wild‐type and pathological Atx3 self‐assembly, we asked if it would affect Atx3 amyloid fibril formation. Under our experimental conditions, the Atx3 variant carrying an expanded polyQ tract (Atx3 77Q) undergoes the same amyloid self‐assembly pathway as the wild‐type protein (Atx3 13Q),^[^ 23, 51 ^]^ as evidenced by the comparable thioflavin T (ThT) progress curves (Figure 2a,d). The addition of a fivefold or tenfold molar excess of CLR01 reduced the fibril assembly rate (v 50) by threefold to fourfold in comparison to Atx3 13Q or Atx3 77Q alone (Figure 2b,e). For the half‐time of the amyloid growth phase (t 50), although CLR01 consistently extended aggregation time by approximately twofold across different experiments, there were no statistically significant differences (Figure 2c,f). Variations in protein preparations across independent replicates contributed to the large variations in t 50, whereas technical replicates within a preparation were highly consistent (Figure 2a,d). The delayed aggregation rates were also observed for the JD in the presence of CLR01 (Figure S6a, Supporting Information). The negative‐control molecule CLR03,^[^ 28 ^]^ which cannot capture lysine or arginine side chains,^[^ 37, 49, 52, 53 ^]^ had no major effect on Atx3 aggregation kinetics (Figure 2; Figure S6a, Supporting Information). The delay in the early steps of Atx3 self‐assembly in the presence of CLR01 was further corroborated by the results of a photoinduced cross‐linking of unmodified proteins assay (Figure S6b, Supporting Information) and by comparing the time‐dependent distribution of monomer and high molecular weight (HMW) soluble Atx3 species formed in the absence and presence of CLR01 using size‐exclusion chromatography (Figure S6c, Supporting Information). Interestingly, despite the delayed aggregation kinetics, CLR01 increased the ThT fluorescence value at the end of the aggregation of JD and both Atx3 variants. Since ThT is a positively charged molecule that often binds to the fibril surface via electrostatic interactions,^[^ 54, 55 ^]^ this likely results from CLR01 altering ThT binding pockets by reversing lysine/arginine charges, enhancing ThT binding and fluorescence. Notably, other proteins have shown an increased ThT‐fluorescence signal in the presence of CLR01, despite reduced fibril formation and cytotoxicity.^[^ 32, 56 ^]^

*CLR01 binding modulates the Atx3 aggregation pathway, delaying amyloid fibril assembly and decreasing the secondary nucleation rate. a,d) Representative ThT progress curves of 5 µm Atx3 13Q (a) or Atx3 77Q (d) amyloid self‐assembly kinetics in the absence or presence of CLR01 or CLR03. Error bars correspond to the standard deviation of three replicates. b,c) Bar plots show mean ± standard error of the mean (SEM) for (b) Atx3 13Q amyloid formation halftime (t 50) and (c) aggregation rate (v 50) derived from fitting of ThT fluorescence curves.[

99

] Control samples were compared with CLR01 or CLR03‐treated samples at the indicated molar ratios. Individual points represent independent replicates (N = 5; each representing the mean values from 3–5 technical replicates). e,f) Same as (b, c) for Atx3 77Q, with 4 independent replicates. Statistical analysis was performed using a linear mixed‐effects model (LME) for Atx3 13Q (b, c), and repeated‐measures one‐way ANOVA for Atx3 77Q (e,f), followed by Dunnett's multiple comparisons test (Table S2, Supporting information). g) Time‐dependent DLS intensity distributions during the aggregation of 5 µm Atx3 13Q alone or in the presence of 25 µm CLR01 or CLR03. Different colors show different time points (size distributions measured in triplicate). Vertical lines: scattering intensities predicted by a nucleation‐and‐growth model for monomers (thinner lines) and amyloid fibrils (thicker lines) at the end of 48 h incubation (refer to Figure S7 in the Supporting Information, for other time points); h) same as (g) for Atx3 77Q. i) Fitted nucleation‐and‐growth rate constants for Ax3 13Q and Atx3 77Q; error bars correspond to standard errors (Table 1).*

To obtain additional mechanistic insight, the change in the size distribution of self‐assembled species of Atx3 variants during the aggregation assay was analyzed using dynamic light scattering (DLS).^[^ 23, 26 ^]^ First, the kinetics of the size distribution changes were measured for wild‐type and Atx3 77Q (Figure 2g,h). The integrated analysis of both ThT fluorescence and DLS size distribution data has previously shown that the aggregation of both Atx3 variants occurs predominantly by secondary nucleation (rate constant k 2), which is substantially faster than primary nucleation (rate constant *k_n_ *) and fibril elongation (rate constant *k_+_ *).^[^ 16, 26 ^]^ The same nucleation‐and‐growth model was used to describe the kinetic curves and DLS data obtained in the absence or presence of CLR01 (Figure 2i and Table 1; Figure S7, Supporting Information). The fitted parameters correspond to the products *k_n_ *Δµ_0_, k +Δµ_0, and k 2_Δµ_0 (Table 1), where Δµ_0 is the initial supersaturation level (as detailed in the Supporting Information). A simplified numerical fitting procedure was adopted taking into account the ThT fluorescence and DLS data. We found that the presence of CLR01 delayed protein aggregation by inhibiting the step of secondary nucleation, while primary nucleation remains the rate‐limiting step. Without CLR01, fibril elongation rates were too low to be quantified due to the predominance of secondary nucleation (*k_+_

• << k 2). For this reason, it is not clear whether the values of k + obtained in the presence of CLR01 correspond to a modified or unmodified mechanism of fibril elongation. The rate constants *k_n_ *, k 2, and *k_+_
• that fit the DLS data also describe the half‐time coordinates measured in the ThT fluorescence experiment (Supporting Information), further supporting the validity of our theoretical analysis. Unlike CLR01, CLR03 did not induce any evident shift in the aggregation mechanism of Atx3 13Q or Atx3 77Q (Table 1).

When examining the aggregate morphologies using transmission electron microscopy (TEM), it became evident that CLR01 did not inhibit Atx3 aggregation but instead reshaped its fibrillar assembly pathway. As depicted in Figure 3a, at the endpoint of the aggregation assay, when the ThT fluorescence signal reaches a plateau for both Atx3 variants (Figure 2a,d), Atx3 13Q formed curvilinear protofibrils with a heterogeneous distribution of lengths, whereas Atx3 77Q showed a mixture of protofibrils and fibrillar bundles corresponding to mature fibrils. In the presence of CLR01, mostly spherical oligomers and small protofibrils were visible at the equivalent time points, and no mature fibrils were observed (Figure 3a). The size distributions of aggregates in each condition (Figure 3b,c) indicate that Atx3 13Q and Atx3 77Q aggregates have a median size of ≈30–40 nm, a characteristic that remains unaltered upon incubation with CLR03. However, when incubated with CLR01, a more uniform population of aggregates forms, with only 12–13 nm in size. This finding aligns with the observation of a population of very small particles with R H = 10 nm in the DLS experiments for both Atx3 variants treated with CLR01 over 72 h (Figure 2g,h; Figure S7a,b, Supporting Information). Because DLS tends to favor detecting larger particles, the prominence of particles with a smaller hydrodynamic radius indicates their high abundance within the sample, as clearly supported by TEM images and further confirmed in the size frequency distribution histogram (Figure 3b). The minor population of Atx3 particles with R H ≈ 150 nm detected by DLS) in the presence of CLR01 is confirmed by the observation of oligomer/protofibril clusters, particularly in the TEM images of Atx3 77Q species formed after prolonged incubation periods of 140 h (Figure S7c,d, Supporting Information), notwithstanding the continued presence of a majority population of spherical oligomers (Figure S8a–c, Supporting Information). The reduction of fibrillar bundles of Atx3 77Q by CLR01 (Figure 3a; Figure S7c, Supporting Information) was supported by the decrease in mature SDS‐resistant fibrils observed in a filter retardation assay (Figure S7e, Supporting Information). Interestingly, high molar ratios of CLR01 were also able to disrupt preformed protofibrils of both Atx3 variants (Figure 3d) and mature fibrils of Atx3 77Q (Figure 3e; Figure S8d, Supporting Information), aligning well with the previous observation of CLR01's ability to disrupt α‐synuclein, Αβ40, Αβ42, and SOD1 fibrils, among others.^[^ 31, 36, 39 ^]^

CLR01 reduces the assembly of mature fibrils of mutant Atx3 and dissociates preformed protofibrils and mature fibrils. a) Morphological analysis of 5 µm Atx3 13Q or Atx3 77Q aggregates in the absence or presence of 50 µm CLR01 or CLR03 at the endpoint of the aggregation assay (66 h), monitored by negative staining TEM. Scale bars correspond to 100 nm; orange rectangles denote Atx3 curvilinear protofibrils; red rectangles highlight small protofibrils formed in the presence of CLR01, red triangles point to Atx3 spherical oligomers; white arrowheads mark clusters of spherical aggregates; and white open arrows point to mature Atx3 Q77 fibrils. b) Particle size histograms and c) scatter plots of all particle measurements with median ± 95% CI, from the TEM images shown in (a) (3–4 images per condition were analyzed). Descriptive statistics are shown in Table S3 (Supporting information). d) Morphological analysis of 5 µm Atx3 13Q or Atx3 77Q protofibrils formed at the endpoint of the aggregation assay (t = 66 h), before (T0) and after three days of incubation (T3d) at 37 °C in the absence (Control) or presence of 400 µm CLR01 or CLR03 by negative‐staining TEM. Scale bars correspond to 100 nm. e) Morphological analysis of 5 µm 77Q mature SDS‐resistant fibrils from a 160 h aggregation assay before (T0) and after ten days of incubation (T10d) at 37 °C in the absence (Control) and presence of 400 µm CLR01 by negative‐staining TEM. Scale bars correspond to 200 nm.

Overall, these results show that CLR01 modulates the early steps of Atx3 self‐assembly, including the disease‐associated Atx3 77Q variant. This effect is achieved by delaying Atx3 self‐assembly and the formation of protofibrils and mature fibrils through a mechanism involving a decrease in the rates of secondary nucleation. Furthermore, CLR01 can dissociate preformed fibril agglomerates, showcasing a multifaceted mode of action that unveils its potential for therapeutic applications in SCA3.

CLR01 Protects Cortical Neurons from Pathological Atx3‐Induced Synapse Decline

2.4

Given CLR01's inhibition of Atx3 self‐assembly, we asked whether it could rescue the synaptic pathology in cortical neurons expressing Atx3 containing an expanded polyQ. Several studies have demonstrated that CLR01 is internalized when added to cultured neurons and is not toxic to the neurons.^[^ 38, 41, 57, 58 ^]^ First, we investigated the effect of CLR01 on the pathogenic impact of polyQ‐expanded Atx3 expression in excitatory synapses of cortical neurons. Cells were transfected with Atx3 constructs containing 28Q (wild‐type) or 84Q (pathological‐disease‐associated variant) on day in vitro (DIV) 9–10 and fixed on DIV 16. In line with previous findings,^[^ 59 ^]^ the expression of the disease‐associated Atx3 84Q variant led to a decrease in the levels of excitatory synapses, evident by the reduction in the intensity of the postsynaptic marker PSD95 colocalized with the presynaptic marker vGLUT1 when compared to cells expressing wild‐type Atx3 (Figure 4a,b).

CLR01 reverts the loss of glutamatergic synapses in cortical neurons caused by pathological Atx3. Rat cortical neurons were transfected with plasmids encoding eGFP, wild‐type eGFP‐Atx3 28Q, or mutant eGFP‐Atx3 84Q. a) Neurons were immunolabeled for MAP2, PSD95, and vGLUT1. Excitatory synapses were detected as PSD95 puncta that colocalized with vGLUT1; scale bar = 10 µm. b) Integrated fluorescence intensity of PSD95 puncta colocalized with vGLUT1 (n = 33–36 neurons per condition in 3 independent experiments). Statistical analyses were performed using Kruskal–Wallis and Dunn's post‐hoc test (Table S4, Supporting Information). c) Neurons incubated with 10 µm CLR03 or CLR01 were immunolabeled for MAP2, PSD95, and VGluT1 clusters; scale bar = 10 µm. d) Integrated fluorescence intensity of PSD95 puncta colocalized with VGluT1 (n = 35–41 neurons per condition in each of 3 independent experiments. Statistical analyses were performed using Kruskal–Wallis and Dunn's post‐hoc tests (Table S5, Supporting Information). In panels (c) and (d), boxes show the 25th and 75th percentiles, whiskers range from the minimum to the maximum values, and the horizontal line shows the median value. Mean is represented by “+.” e) GFP‐Atx3 accumulates in the cell body of a fraction of transfected cortical cultures treated with CLR03 or CLR01; scale bars = 300, 15 µm. f) Percentage of cells with GFP‐Atx3 aggregates (n = 3 independent experiments, 31–35 neurons per condition). Statistical analyses were performed using two‐way ANOVA and Sidak's multiple comparisons post‐hoc tests (Table S6, Supporting Information). p‐values are indicated in the graphs shown in panels (b), (d), and (f). Data are represented as mean ± SEM.

Incubation with CLR01 for 24 h rescued the PSD95/vGLUT1 colocalization levels, whereas the negative‐control compound CLR03 did not (Figure 4c,d). Importantly, CLR01 incubation did not change the labeling for PSD95 in control neurons (Figure S9a,b, Supporting Information). Next, we assessed whether the reversal of synaptic pathology induced by CLR01 was linked to a decrease in Atx3 aggregation within neurons. To identify Atx3 aggregates, cortical neurons transfected with eGFP‐tagged Atx3 84Q were examined for eGFP‐signal accumulation in the absence or presence of CLR01 or CLR03. Cells were categorized as displaying eGFP‐positive aggregates if they exhibited at least one observable eGFP‐positive accumulation, regardless of size. Cells showing a diffuse eGFP signal were classified as lacking aggregates. The results showed no changes in the number of cells containing Atx3 aggregates upon incubation with CLR01 (Figure 4e,f). Based on the results shown above with the recombinant protein, CLR01 does not abolish aggregation but traps aggregates into small, spherical oligomers that persist for long incubation times. Meanwhile, large aggregates still form, and the shorter protofibrils (30–60 nm) are markedly reduced. Therefore, synaptic improvement may result from the selective reduction of these protofibrils, while the larger aggregate species are likely to be less harmful, in line with prior studies describing the differential synaptotoxicity of Aβ aggregate conformers.^[^ 60, 61 ^]^

In summary, these results show that CLR01 decreased the excitatory synaptic protein content in cortical neurons expressing pathological Atx3. Interestingly, this effect was not due to a reduction in the number of cells with microscopically visible Atx3 aggregates, suggesting that these species do not impact synapse integrity.

CLR01 Treatment Reduces Neurotoxicity in a C. elegans Model of SCA3

2.5

To evaluate if CLR01 could ameliorate disease‐related phenotypes in vivo, we used a C. elegans model of SCA3, which expresses human Atx3 protein with a polyQ tract length of 130Q in the C. elegans nervous system and exhibits age‐related locomotion defects, mimicking key features of the disease.^[^ 62 ^]^ First, a food clearance assay was used to identify nontoxic concentrations of CLR01 in wild‐type nematodes. None of the CLR01 concentrations tested (0.0001–100 µm) had an impact on the growth or survival of the animals after six days of exposure (Figure S10, Supporting Information). Next, we tested the effect of chronic presymptomatic treatment with the same CLR01 concentrations in the C. elegans model of SCA3. As illustrated in Figure 5a, treatment with CLR01 significantly improved the animals' motor impairments, 0.1 µm being the most effective dosage (Tables S8, S10, Supporting Information). Importantly, treatment with a similar concentration of CLR01 had no impact on the motility of wild‐type animals, showing the specificity of the effect (Figure 5b).

*CLR01 treatment alleviates motor deficits in a C. elegans model of SCA3 even when treatment is initiated at a postsymptomatic disease phase. a) CLR01 treatment improved motor dysfunction of mutant Atx3 animals. AT3q130 animals were treated with different concentrations of CLR01 for four days, and the percentage of locomotion‐defective animals was determined. CLR01 significantly impacted motor phenotype between 0.0001 and 1 µm; the maximum efficiency (%) was observed at 0.1 µm at which 67% phenotypic rescue was achieved compared to the control strain (Table S8, Supporting Information). b) Treatment with 0.1 µm CLR01 for four days did not impact motor phenotype in the wild‐type strain (N2). c) Postsymptomatic treatment with CLR01 improves the animals' motor defects after 2 or 4 days of treatment (days 6 and 8 after hatching). (a–c) Data are represented as mean ± SD of 4 independent experiments, with ≈50 animals per condition/per assay (total number of animals of ≈250). A one‐way ANOVA test was applied, followed by a bilateral Dunnet test for post‐hoc comparisons (a and c). In the comparison between the two groups (b and c before treatment), independent‐samples t‐test was used (Tables S9, S10, Supporting Information). d–g) Impact of early life chronic CLR01 treatment on mutant human Atx3 aggregation pattern, labeled with an Atx3‐specific antibody in the C. elegans model of SCA3 throughout animals’ development and adulthood. Pictures of the animal's heads were obtained by confocal microscopy (Olympus FV1000) and analyzed using the MeVisLab software.[

106

] The graphs show a pool of 3 independent assays, with at least four animals per condition in each trial. p‐values were calculated using independent‐samples t‐test (Table S10, Supporting Information). Scale bars = 20 µm.*

Since the most frequent clinical situation in SCA3 is symptoms‐driven genetic testing and diagnosis, we subsequently tested whether CLR01 administration would also be effective when treatment is initiated after the onset of motor impairment. Notably, even when treatment was initiated after the symptoms’ onset, CLR01 still reduced the animals’ motor dysfunction, observed by a partial reduction of the motility impairments after two days of treatment. This decreased disease progression and severity in the animals, as shown in Figure 5c (Table S9, Supporting Information). The reversal of SCA3‐related symptoms in vivo by CLR01, even upon postsymptomatic treatment, is consistent with its ability to disrupt preformed fibrillar bundles of pathological Atx3 (Figure 3c,d). In agreement with in vitro ThT and TEM data (Figure 3a–c) and with the results obtained in cultured cortical neurons (Figure 4e,f), CLR01 treatment did not impede the aggregation of pathological Atx3 in SCA3 animals (Figure 5d–g; Table S10, Supporting Information) but improved disease‐related motor impairment in the C. elegans SCA3 model, suggesting that the Atx3 aggregates formed upon treatment with CLR01 have reduced toxicity. Fluorescent aggregated foci containing Atx3 were observed within 24 h and remained unchanged by CLR01 treatment up to animal adulthood. Interestingly, the number of aggregates formed during the treatment did not significantly change after two days. However, the size of the aggregate clusters temporarily increased before decreasing again during the treatment. These results are in line with CLR01's ability to reduce secondary nucleation rates and, in the long term, also disrupt amyloid fibrils in vitro, making it a promising candidate for further investigation and preclinical testing.

CLR01 Treatment Delays Disease Onset and Improves Motor Deficits in a SCA3 Mouse Model

2.6

Next, we asked if CLR01 treatment could ameliorate disease‐related phenotypes in a mammalian SCA3 model. Taking advantage of the CMVMJD135 transgenic mouse model of SCA3 (SCA3 mice), in which quantifiable features recapitulating the motor symptoms observed in SCA3 patients can be assessed,^[^ 6 ^]^ we conducted preclinical experiments testing the therapeutic effects of CLR01, which was previously shown to be able to cross the blood–brain barrier and be internalized by neuronal cells.^[^ 41, 58 ^]^ This mouse model is well‐established and has been widely used to test candidate therapies for SCA3. It presents several behavioral and neuropathological traits that closely mimic the human condition. Because CAG repeat instability is observed in the SCA3 mouse, similar to that observed in humans,^[^ 63 ^]^ it is important to control the number of CAG repeats within the different experimental groups. Thus, we verified that no differences existed between the mean CAG repeat number in SCA3‐vehicle‐ and CLR01‐treated groups (Figure S11a, Supporting Information), allowing us to avoid CAG repeat number as a confounder.^[^ 64 ^]^ In CMVMJD135 mice, the onset and progression of the symptoms occur gradually during the life of the animal, the first symptoms appearing at week 6, manifesting as loss of limb strength, followed by other motor deficits that worsen with age until the motor phenotype is fully established by week 16. The neuropathological findings, including mutant Atx3 nuclear inclusions and neuronal loss, occur later in life, around 22 weeks, and progress over time. In this study, CLR01 treatment was initiated early in life, at four weeks, when neurological symptoms are not yet detected in the SCA3 mice, and lasted 18 weeks (Figure S11b, Supporting Information).^[^ 6, 8 ^]^

Chronic treatment with CLR01 caused no detrimental effects to WT mice, as their body weight (Figure S11c, Supporting Information), fur quality, activity, presence of grooming, and nesting behaviors were comparable to those of wild‐type mice receiving vehicle. CLR01 treatment delayed the reduction of body weight seen in the SCA3 mice, a two‐week delay being observed; nevertheless, at 12 weeks of age, the body weight loss reached the SCA3 vehicle group level, suggesting that this effect on body weight was not sustained throughout time (Figure S11c, Supporting Information). No evident toxicity was observed, as CLR01‐treated mice performed similarly to the vehicle‐treated mice in all the motor tests, suggesting a safe profile for CLR01 chronic treatment in these mice, similar to previous models.^[^ 31 ^]^ CLR01 treatment showed a strong therapeutic effect on early motor symptoms in the SCA3 mice, including a delay in the symptom onset and overall improved performance in the swimming test (Figure 6a) and to a lesser extent in the balance‐beam test, in which at 12 weeks of age the CLR01‐treated SCA3 mice performed indistinguishably from WT mice, but by 16 weeks of age, their performance deteriorated and was similar to that of vehicle‐treated SCA3 mice (Figure 6b). Importantly, CLR01 administration substantially improved the gait quality of SCA3 mice and delayed the appearance of an abnormal gait pattern by eight weeks (Figure 6c).

CLR01 chronic administration improves motor function and neuropathology of SCA3 mice. a) The latency to cross a swimming tank, representing both the strength and the movement coordination ability of the animals, is higher in vehicle‐treated SCA3 mice. The severity of this symptom improves over time with CLR01 treatment, which has a delayed onset and is effective up to 16 weeks of age. b) CLR01 improves SCA3 mice balance, decreasing their latency to cross a 12 mm squared beam. c) The gait quality of SCA3 animals is strikingly improved by CLR01 treatment, delaying the onset of ataxic gait by eight weeks. d) Representative images of cresyl violet staining in the cervical spinal cord of healthy and pyknotic cells. e) Vehicle‐treated SCA3 mice showed a substantial increase in the number of pyknotic cells, which was rescued by daily CLR01 administration for 18 weeks. f) Representative pictures of the motor‐neuron cell bodies (black arrows) in the ventral horn of the cervical spinal cord. g) CLR01 treatment increases the number of motor neurons in the cervical spinal cord. Sample size and statistical details are provided in Table S11 (Supporting Information). The scale bar for lower magnification pictures represents 200 µm, and for higher magnification, 100 µm. p‐values are shown in the graph. (a, b) Data are represented as mean ± SEM (n = 13–19 per group). A repeated measures ANOVA test was applied, followed by Tukey test for multiple comparisons. (c) Data are represented as the % of animals within each category (n = 13–19 per group). A Kruskal–Wallis H test was applied. (e and g) Data are represented as the median (min; max) with all datapoints shown (N = 5–6 animals per group; n = 4–7 slices per mouse). A one‐way ANOVA test was applied, followed by a Tukey test for multiple comparisons. Black lettering represents the comparisons between WT‐vehicle and SCA3‐vehicle. Red lettering represents the comparisons between SCA3‐vehicle and SCA3‐CLR01.

Next, we asked if the neuropathology features of SCA3 mice were improved by CLR01 treatment. We have shown previously that pyknosis, a process in which the cell nucleus becomes dense and compact, indicating the initiation of a degeneration process, occurs in the brain of the SCA3 mice.^[^ 6, 8 ^]^ CLR01 chronic treatment fully rescued this degenerative process in the cervical spinal cord of SCA3 mice by reducing the number of pyknotic cells to vehicle‐treated WT mice levels (Figure 6d,e). Previous studies have shown that at 34 weeks of age, cholinergic motor neurons degenerate in the spinal cord of SCA3 mice.^[^ 6, 8 ^]^ Interestingly, CLR01 treatment led to the preservation of cervical spinal cord neurons in SCA3 mice and did not impact the number of these neurons in WT animals (Figure 6f,g). Additionally, CLR01 administration had no major impact on the number of Atx3‐positive nuclear inclusions (Figure S12, Supporting Information), which are usually found in the brain and spinal cord of SCA3 mice.^[^ 6, 8, 65 ^]^ As the presence of these end‐stage aggregates in CLR01‐treated mice is associated with improved histopathological findings, we propose that the detected Atx3 aggregated species have a reduced proteotoxicity.

Discussion

3

Protein misfolding and its subsequent assembly into amyloid and amyloid‐like deposits represent distinctive features consistently observed in numerous severe neurodegenerative conditions, such as Alzheimer's, Parkinson's, and Huntington's diseases. In SCA3, similar to other polyglutamine expansion disorders, intranuclear inclusions immunopositive for Atx3 are found in neurons.^[^ 4, 66, 67 ^]^ Although the exact role of Atx3 self‐assembled molecules in causing selective neuronal demise has not yet been fully deciphered,^[^ 68, 69 ^]^ several approaches to reduce the aggregate burden have been pursued as strategies for the development of disease‐modifying therapies for SCA3. Here, we show that CLR01, a supramolecular ligand with an affinity for lysine and, to a lesser extent, arginine residues,^[^ 31, 70 ^]^ modulates the aggregation behavior of Atx3 in vitro and positively influences the reversal of disease‐related phenotypes in SCA3 cell and C. elegans models, as well as delaying the onset of symptoms in a murine SCA3 model (Figure 7). These effects are observed without a discernible reduction in neuronal aggregates. In line with the observation that CLR01 stabilizes spherical oligomers without entirely inhibiting the formation of large aggregates, we propose that these microscopically visible aggregates formed in the presence of CLR01 are unlikely to be toxic. Instead, the population of heterogeneously sized protofibrils, which are mainly formed through secondary nucleation processes and significantly decreased by CLR01, is possibly the most toxic entity.

Schematic representation of the effects of CLR01 on Atx3 self‐association. Binding of CLR01 to K128 (colored dark blue) within a positively charged surface patch opposite to the aggregation‐prone region and ubiquitin binding site (orange) triggers a conformational change in the helical hairpin (yellow) of the Josephin domain. The CLR01‐induced conformational change reshapes the Atx3 aggregation pathway and traps spherical oligomers while reducing secondary nucleation rates. This modulation of the self‐assembly pathway correlates with the reversion of synapse dysfunction in SCA3 cell models and a delay in the onset of motor impairment in SCA3 animal models. Image created in BioRender (https://BioRender.com/lgwm68m).

An Optimal Chemical Environment Drives CLR01 Recruitment to the Josephin Domain

3.1

In silico and in vitro studies showed that CLR01 selectively binds to the globular JD, a key player in initiating the early stages of fibrillar aggregate formation. To explore the interaction mechanism, we combined NMR, which is highly sensitive to changes in the local chemical environment of the ligand binding sites, with extensive enhanced sampling MD simulations that can capture transient interactions while sampling various conformational states. We showed that CLR01 forms a complex with K128, located in a positively charged region in the vicinity of an exposed hydrophobic surface patch (Figure 7). This region displays an ideal chemical environment for the establishment of favorable interactions with CLR01. While encapsulating K128, the phosphate groups of the molecular tweezer can engage in electrostatic interactions, and its convex surface may be further stabilized through hydrophobic interactions with neighboring hydrophobic residues. Upon binding to K128, CLR01 modulates the conformational fluctuations in the distal helical hairpin through allosteric long‐range effects that restrict the local misfolding necessary for exposing the amyloidogenic region in helix α4, which initiates the self‐assembly of both wild‐type and pathological Atx3 variants. CLR01 binding to the N‐terminal region (N17) of huntingtin exon‐1 has also been shown to induce conformational changes that reduce its aggregation.^[^ 37 ^]^ It has been suggested that CLR01 likely interferes with the assembly of a dimer of tetramers, a key initial aggregation step involving the N17 sequence, which subsequently nucleates the formation of polyQ amyloid fibrils.^[^ 71 ^]^ Importantly, although modulating abnormal protein self‐assembly and facilitating degradation of toxic assemblies have been proposed previously as the main mechanisms of action of CLR01,^[^ 72 ^]^ this study reports for the first time an allosteric long‐range inhibitory effect as the mechanism underlying CLR01's therapeutic activity.

CLR01 Decreases Secondary Nucleation Processes, Delays Atx3 Fibril Assembly, and Dissociates Preformed Amyloid Fibrils

3.2

The analysis of Atx3 aggregation in the presence of CLR01 revealed a pronounced kinetic impact on amyloid fibril assembly. Our in vitro assays confirmed that CLR01 effectively remodels the assembly of Atx3 into amyloid structures and delays, but does not prevent, the formation of mature fibrils of the pathological Atx3 77Q. We have shown previously that surface‐mediated secondary nucleation prevails in the in vitro kinetics of Atx3 fibril formation of both wild‐type and disease‐related variants.^[^ 16, 26, 27 ^]^ CLR01 specifically decreases the rate of secondary nucleation and thereby restricts the autocatalytic amplification of Atx3 amyloid species. Consistently, incubation with CLR01 reduces the formation of medium‐sized protofibrils and entraps the Atx3 aggregates into small quasispherical oligomers, in clear contrast to the aggregate populations formed in the absence of the molecular tweezer (Figure 7). Several scenarios can be put forward to explain the impact of CLR01 on these secondary nucleation processes. One possibility is that CLR01 interaction may restrict the local misfolding and the ability of Atx3 monomers to adopt the structure required for nucleation and assembly at the surface of preformed fibrils. Alternatively, it may directly impede the binding of monomers/oligomers to the fibrils or modify the properties of the parent fibrils required for the attachment step. It remains to be determined whether the population of spherical oligomers that predominate during incubation with CLR01 are on‐pathway amyloid species or structurally distinct Atx3 aggregate polymorphs that may later convert into extended fibrillar aggregates with higher thermodynamic stability.^[^ 73 ^]^ In view of the inhibition of secondary nucleation by CLR01, it is tempting to hypothesize that the fibrils formed in the presence of the compound have a distinct conformational arrangement of Atx3 and form a different strain or polymorph. Future studies addressing a time‐resolved structural analysis of fibril polymorphs formed along the aggregation pathway, similar to the recent structural analysis of tau^[^ 74 ^]^ or human islet amyloid polypeptide^[^ 73 ^]^ will help to address these questions. Similarly, it remains unclear from our current data whether the structural characteristics of the larger Atx3 aggregate polymorphs formed after prolonged incubation with CLR01 differ from those formed in the absence of the molecular tweezer. It is also noteworthy that CLR01 dissociated preformed Atx3 protofibrils and mature fibrils, exhibiting a fibril‐disrupting activity similar to those shown for α‐synuclein^[^ 36 ^]^ and Aβ^[^ 57 ^]^ aggregates. This is a slow process that requires a high molar excess of the molecular tweezer, suggesting that CLR01 could be used as a treatment not only in the early but also at later stages of the disease. The exact mechanisms that lead to the disruption of the Atx3 fibrils have yet to be fully understood. However, we speculate that at relatively high molar concentrations, CLR01 may interact with weaker affinity binding sites containing exposed positively charged residues in the C‐terminal tail of Atx3 and weaken the intermolecular interactions that stabilize the fibrillar assemblies.

CLR01 Reverts SCA3‐Related Phenotypes in Cell and Animal Disease Models

3.3

The thorough mechanistic analysis of CLR01's impact on Atx3 aggregation, particularly its role in reducing secondary nucleation, a process closely associated with neurotoxicity and the proliferation and spreading of amyloid fibrils,^[^ 75, 76 ^]^ prompted the investigation into its effects in SCA3 cell and animal models. One key question is how the mechanisms for aggregation impairment dissected in vitro translate into tangible improvements of the pathology in vivo. Importantly, in our SCA3 neuronal model, CLR01 treatment rescued the synapse pathology occurring when cultured cortical neurons express pathological Atx3 with an expanded polyQ tract, although our data do not permit us to infer if these changes correlate with neuronal survival. Using a similar model, we have previously shown that the expression of expanded ataxin‐3 decreases dendritic complexity relative to the nonexpanded variant.^[^ 59 ^]^ This complementary analysis could be considered in future studies to further elucidate the impact of CLR01 on synaptic function. Importantly, CLR01 restored motor dysfunction in the SCA3 C. elegans model and showed a clear therapeutic effect in SCA3 mice. In this context, CLR01 delayed the onset of motor symptoms and rescued spinal cord and motor‐neuron degeneration. Importantly, although survival was not assessed in this study, it represents a crucial measure of translational relevance that should be considered in future research.

It is noteworthy that the efficacy of CLR01 in reverting or delaying SCA3‐related phenotypes in cultured cortical neurons and in vivo was not accompanied by a reduction in microscopically visible Atx3 aggregates. In fact, the identification of Atx3 deposits in both affected and nonaffected brain regions of patients who died of SCA3 has suggested that these microscopically visible inclusions may not represent the neurotoxic entities.^[^ 68, 69 ^]^ As in other amyloid diseases, growing evidence supports the idea that oligomeric and/or protofibrillar species play a critical role in triggering and spreading polyQ pathology in neurons.^[^ 77, 78, 79 ^]^ The beneficial effects observed combined with the presence of large aggregates in SCA3 cells and animal models suggest that in cultured neurons and in vivo, CLR01 may indeed interfere with secondary nucleation processes responsible for the autocatalytic proliferation and spreading of toxic Atx3 protofibrils, which are stealth to the applied imaging methodologies. In support of this hypothesis, treating SCA3 mice with CLR01 from an early disease state protected the neuronal dysfunction, enabling their preservation even after symptom onset. Combining fine‐tuned cytotoxicity assays with structural characterization of the Atx3 aggregates will be essential to clarify whether larger aggregates formed in the presence of CLR01 are nontoxic, and to further investigate the role of smaller, transient protofibrillar species, which our findings suggest may be the most toxic forms.

The Temporary Efficacy of CLR01 Treatment in the SCA3 Mouse Model

3.4

An intriguing observation in the mouse model is that despite a delayed onset of several disease‐related motor symptoms, once the symptoms appear, they evolve rather quickly to levels that mirror those of the untreated SCA3 mice. Several hypotheses may be put forward to explain this observation. First, one may argue that CLR01 reshapes the Atx3 aggregation pathways, leading first to the appearance of nontoxic Atx3 amyloid intermediates, which are kinetically favored. Later in the aggregation process, these may convert to thermodynamically more stable (proto)fibril polymorphs that are competent to template the assembly of new fibrils, which grow exponentially, accumulate, and spread in the mouse brain and spinal cord, resulting in motor dysfunction. Alternatively, the continuous subcutaneous administration of CLR01 could result in the accumulation of the molecular tweezer in the mouse brain.^[^ 38 ^]^ This build‐up may drive CLR01 interactions with weaker affinity binding sites on Atx3, including lysine residues necessary for Atx3 posttranslational modifications^[^ 80, 81, 82, 83 ^]^ and nucleocytoplasmic shuttling,^[^ 84, 85 ^]^ with implications on protein stability aggregation and turnover.^[^ 86, 87, 88 ^]^ If this is the case, less‐frequent administration of CLR01, e.g., 2–3 times a week, as has been done in recent studies,^[^ 41, 89 ^]^ may have a better therapeutic effect than the daily administration used here. The interaction of CLR01 with lysines/arginines on secondary binding sites might also explain the decreased effect of CLR01 treatment on locomotion defects in C. elegans observed at concentrations higher than 1 µm. The transient beneficial effects of CLR01 treatment in the mouse model suggest that the molecular‐level mechanistic information gathered in this study will be critical for developing improved lead molecules for effective SCA3 therapies.

CLR01 as a Scaffold for Allosteric Modulators of Pathological Atx3 Aggregation

3.5

SCA3 is a severe neurodegenerative disease with an unmet and urgent need for disease‐modifying therapies. Numerous studies have demonstrated the efficacy of therapeutic strategies, such as silencing the pathological polyQ‐expanded Atx3 or addressing downstream proteotoxicity.^[^ 90, 91 ^]^ We have recently shown that increasing dopamine levels offers a promising approach to stopping disease progression by interfering with the secondary nucleation step of Atx3 aggregation.^[^ 16 ^]^ However, the lack of a comprehensive analysis of their mechanisms of action has hindered the translation of these findings into clinical applications. This study presents a systematic, multiscale analysis of CLR01's efficacy and safety, spanning from the molecular level to the SCA3 mouse model. Our results demonstrate that CLR01 targets the globular domain of Atx3, limiting the exposure of the aggregation‐prone site, which is crucial for initiating early aggregation steps. Furthermore, CLR01 interferes with the autocatalytic proliferation of Atx3 fibrils through secondary nucleation, an effect that is translated into the overall improvement of SCA3 phenotypes in cultured neurons and animal models. Notably, CLR01 delays the onset of motor function loss in a preclinical SCA3 mouse model, hinting at its potential as a disease‐modifying therapy for SCA3 patients, which could be used in combination with other therapeutic approaches currently under development. Our results offer crucial mechanistic insights that advance our comprehension of the self‐assembly pathways of Atx3 in SCA3 and underscore the role of the distal termini of the JD in shaping its conformational landscape and aggregation propensity. A previously unknown allosteric site distant from the ubiquitin substrate binding region was identified, providing critical knowledge for designing tailored molecular tweezers capable of precisely targeting the pathological aggregation routes while preserving the normal function of the native protein in cell proteostasis. These findings provide novel opportunities to identify allosteric binding pockets in “undruggable” aggregation‐prone proteins associated with multiple human diseases, whose self‐assembly regions are highly flexible and challenging to target.

Experimental Section

4

Atx3 Variants and Compounds

Human full‐length Atx3 (Uniprot P54252‐2) containing 13 or 77 glutamines within the polyglutamine tract (Atx3 13Q, MW of 43 582.5 Da or Atx3 77Q, MW of 51 882 Da, respectively), and the truncated variant comprising the N‐terminal globular Josephin domain (Atx3 JD, MW of 23 767.87 Da) were used in the experiments, all containing a N‐terminal 6xHis‐tag. The details about the expression vectors used were in the plasmid repository database Addgene with the accession codes 184247, 184248, and 184249. CLR01^[^ 28 ^]^ and CLR03^[^ 92 ^]^ were provided by Gal Bitan's lab.

Protein Purification

Atx3 variants were expressed in Escherichia coli and purified by metal‐affinity chromatography and size‐exclusion chromatography as described previously.^[^ 23, 26, 51 ^]^ The protein concentration was determined by measuring the absorbance at 280 nm using the extinction coefficient of 35 660 m ^−1^ cm^−1^ (Atx3 JD) or 31 650 m ^−1^ cm^−1^ (Atx3 13Q and Atx3 77Q). The purified proteins were stored at −80 °C in 20 mm sodium phosphate pH 7.5, 150 mm NaCl, 5% v/v glycerol, 2 mm ethylenediaminetetraacetic acid, 1 mm dithiothreitol (DTT).

NMR Titrations

SOFAST‐HMQC NMR spectra were recorded at 21 °C on a 800 MHz Bruker Avance instrument (Bruker, Germany) equipped with a cryogenic probe. Samples of ^15^N‐labeled JD were used at 100 µm after overnight dialysis in a buffer containing 20 mm sodium phosphate pH 7.5, 100 nm NaCl, and 1 mm DTT. 10% D_2_O was added before measurements. CLR01, dissolved in a matching buffer, was titrated into the JD solution to reach a final protein:CLR01 molar ratio of 1:2 in 0.25 increments. The results were analyzed by comparing the chemical shift perturbation effects observed by titration of CLR01 with the spectral assignment of the isolated JD deposited in the BMRB database (entry number 6241).

Enhanced Sampling Simulations

The coordinates of the open (PDB code 1YZB^[^ 46 ^]^) and closed (PDB code 2AGA^[^ 93 ^]^) conformations of the JD were used as the initial coordinates for the GaMD simulations^[^ 42 ^]^ in the absence or the presence of CLR01, as described in the Supporting Information. The CHARMM36m force field^[^ 94 ^]^ was used for the protein in all simulations. The CHARMM‐GUI server^[^ 95, 96 ^]^ was employed for the preparation of all systems. During the GaMD protocol, 2.4 ns of classical MD were carried out, followed by 8–12 ns of GaMD equilibration until convergence of k0D and k0P to 1. Then, dual boost GaMD production runs were carried out (3 × 200 ns replicas) for all the setups using AMBER 20.^[^ 97 ^]^ In addition, JD (open conformation) was simulated in complex with a single CLR01 molecule positioned at K128, with three replicas of 500 ns each. The CL‐FEP approach^[^ 48 ^]^ was employed to estimate the Gibbs energy of binding of CLR01 to K/R residues of the JD, as described in the Supporting Information.

Native Mass Spectrometry

Protein solutions were buffer‐exchanged into 20 mm ammonium acetate using 10 kDa Amicon filters and diluted to 10 µm. CLR01 was added to the solution at 10 µm to induce binding. Solutions containing protein and protein/CLR01 complexes were sprayed via nanospray capillaries (ThermoFisher Scientific) on a Synapt G2‐Si (Waters). The instrument was run in mobility TOF mode with a capillary voltage of 1–2 kV. The sampling cone was set to 20 V and the source offset was set to 20 V. Mobility was measured over the entire m/z range and mobiligrams were extracted from the [M+nH]* ^n^ * ^+^ peak. Mass spectra were deconvoluted using UniDec.^[^ 98 ^]^ Key UniDec parameters included limiting the m/z range to 500–4000 m/z, using background subtraction, and sampling the mass every 1 Da. The relative abundance values provided by UniDec were inserted into Equation (1) to calculate K D values.

$$[eqn]$$

Atx3 Aggregation Assays and TEM Image Analysis—ThT Binding Assay

Immediately before each aggregation assay, purified protein solutions were applied to a Superose 12 10/300 GL column (GE Healthcare Life Sciences) pre‐equilibrated in aggregation buffer (20 mm N‐(2‐hydroxyethyl)piperazine‐29‐(2‐ethane‐sulfonic acid) pH 7.5, 150 mm NaCl, 1 mm DTT) and 0.4 mL fractions were collected. The peak corresponding to monomeric protein was collected for further aggregation assays. ThT assays were carried out in the absence or presence of CLR01 or CLR03 at 37 °C, as described previously, in low protein‐binding Corning Thermowell 96‐well polycarbonate polymerase chain reaction microplates. A protein concentration of 5 µm and ThT concentration of 30 µm were used with 4–5 different biological replicates with 3–5 technical replicates each.^[^ 23, 26, 51 ^]^ For each biological replicate, control and 3 conditions (CLR01 (1:5); CLR01 (1:10); and CLR03 (1:10)), were measured in parallel (paired experiments). The fluorescence (emission 480 nm, excitation 440 nm) was monitored every 30 min on a FluoDia T70 microplate fluorimeter (Photon Technology International, Edison, NJ, USA). Each replicate was fitted to a nucleation and growth model to determine the kinetic parameters t 50 and v 50.^[^ 99 ^]^ The mean from the technical replicates for each biological replicate and tested condition was then calculated. Data were analyzed using Prism 10 (GraphPad Software).

Atx3 aggregate morphologies were imaged at several time points during the aggregation assay by TEM using a TEM JEM‐1400 (JEOL, Tokyo, Japan) at an accelerating voltage of 80 kV, as described previously.^[^ 23, 26, 51 ^]^ Size distributions of aggregates present in TEM images were determined using the open‐source ImageJ Fiji software. Electron microscopy images were calibrated, cropped to remove scale bars, contrast‐enhanced, treated using the Gaussian Blur filter with sigma = 1.5 and the Find Edges option, and binarized using 80–255 or 85–255 threshold. Then, particle analysis was performed differently for round oligomers and elongated fibrils. In the first case, a minimum area size of 40 nm^2^ and a minimum circularity of 0.90 were set before analysis. In the case of fibrils, a first analysis was performed setting a maximum circularity of 0.20–0.4 and then the Shape Filter plugin was used to select particles with a minimum aspect ratio of 2.7.^[^ 100 ^]^ Oligomer and fibril lengths were approximated to the maximum Feret diameter, i.e., the longest distance between any two points along the particle's boundary.

Atx3 Aggregation Assays and TEM Image Analysis—Atx3 Oligomerization Time‐Course Analysis

The aggregation of the proteins in the absence or in the presence of CLR01 or CLR03 was evaluated by size exclusion chromatography (SEC), TEM, and DLS. The proteins (5 µm final concentration) were incubated in aggregation buffer at 37 °C without shaking. At different time points, 50 µL aliquots were collected and analyzed directly by TEM or filtered and injected onto a Superdex 200 Increase 5/150 GL column (GE Healthcare Life Sciences) as described previously for Atx3 13Q.^[^ 16, 26 ^]^ Measurements of the soluble species present in each sample (monomers, intermediate oligomeric species, HMW oligomers) were made by determining the relative peak areas at each time point after integration of SEC chromatograms peak areas and normalization by the total area of the peaks. Measurements of DLS autocorrelation functions were carried out on a Zetasizer Nano ZS DLS instrument (Malvern Instruments) during the aggregation of 5 µm protein in the absence or presence of fivefold or tenfold molar excess of CLR01 or CLR03, as described previously.^[^ 16 ^]^ The samples were incubated in a UV‐Cuvette microcuvette (Brand, Germany) with a cap, maintained at a constant temperature of 37 °C without shaking, and analyzed periodically at 0, 16, 24, 48, 72, and 120 h at 25 °C and a scattering angle of 173° to the incident beam. Intensity size distributions were obtained by CONTIN analysis of the measured autocorrelation functions using the MATLAB code rilt.^[^ 101, 102 ^]^

Atx3 Aggregation Assays and TEM Image Analysis—Atx3 Fibril Dissociation Assays

At the endpoint of the aggregation assay (t = 66 h), determined by the point at which the ThT fluorescence plateaued in a parallel microplate assay, Atx3 13Q and 77Q protofibrils were incubated either in the absence or presence of an 80‐fold molar excess of CLR01 or CLR03 (37 °C, without shaking). For the evaluation of the dissociation of mature fibrils, Atx3 77Q fibrils grown for ≈160 h were used in the assay. The fibril dissociation profile was assessed via TEM as previously described, following 0, 1, 3, or 10 days of incubation.

Neuronal Cell Model—Primary Cortical Neuronal Cultures

Dissociated neuronal cultures were prepared as described previously.^[^ 103 ^]^ Neurons were plated in neuronal plating medium (MEM supplemented with 10% v/v horse serum, 0.6% w/v glucose, and 1 mm pyruvic acid) onto poly‐d‐lysin‐coated coverslips. Low‐density cultures were plated in culture dishes at a final density of 11.6k or 13.9k cells cm^−2^. After 2–4 h of incubation at 37 °C in a humidified incubator containing an atmosphere of 5% CO_2_/95% air, coverslips were flipped over an astroglial feeder layer (Banker cultures) in Neurobasal medium (NBM; supplemented with SM1 (1:50 dilution), 0.5 mm glutamine, 0.12 mg mL^−1^ gentamycin). The cultures were maintained at 37 °C in a humidified incubator under 5% CO_2_/95% air, and fed once per week with fresh NBM without glutamate. To prevent glia overgrowth, neuronal cultures were treated with 10 µm 5‐deoxyfluoruridine after 3 DIV. All animal procedures were reviewed and approved by ORBEA and Direção Geral de Veterinária (DGAV, Portugal).

Neuronal Cell Model—Transfection Protocol

DNA constructs were transiently expressed in primary neuronal cultures using a calcium phosphate transfection protocol adapted from Jiang et al.^[^ 104 ^]^ Cultures were transfected at 10 DIV, with 3 µg of plasmids encoding eGFP, nonexpanded eGFP‐Atx3 28Q, or expanded GFP‐Atx3 84Q. All incubations were performed at 37 °C in a humidified incubator under 5% CO_2_/95% air. Plasmids were allowed to express for 6 days.

Neuronal Cell Model—Immunofluorescence

Low density cultured neurons (DIV 16) were fixed with 4% v/v paraformaldehyde (PFA)/4% w/v sucrose in phosphate buffer saline (PBS; in mm 137 NaCl, 2.7 KCl, 1.8 KH_2_PO_4_, 10 Na_2_HPO_4_·2H_2_O, pH 7.4) for 15 min at room temperature, followed by 6 sequential washes with PBS. After permeabilization with 0.25% w/v Triton X‐100 in PBS for 5 min at 4 °C, neurons were washed with PBS and blocked with 10% w/v bovine serum albumin (BSA) for 30 min at 37 °C. Subsequently, cells were incubated with primary antibodies diluted in 3% BSA overnight at 4 °C or for 2 h at room temperature. To stain for excitatory synapses, anti‐PSD95 antibody (mouse, 1:500, Cell Signaling Technology) and anti‐VGLUT1 antibody (guinea pig, 1:1000, Merck Millipore) were used; dendrites were identified with anti‐MAP2 antibody (chicken, 1:5000, Abcam). Neurons then were washed 6 times with PBS and reincubated for 1 h with secondary antibodies in 3% w/v BSA at 37 °C. The following secondary antibodies from Invitrogen Molecular Probes were used: Alexa Fluor 488 conjugated anti‐rabbit (1:500), Alexa Fluor 568 conjugated anti‐mouse (1:500), Alexa Fluor 647 conjugated anti‐guinea pig (1:500), and AMCA‐conjugated anti‐chicken (1:200, Jackson ImmunoResearch). Following six washes, coverslips were mounted with fluorescence mounting media.

Neuronal Cell Model—Quantification of Aggregate‐Containing Neurons

Transfected cortical neurons at DIV 16 were fixed as described for immunofluorescence. 10–13 neurons per condition were visualized using a Carl Zeiss Axio Observer Z1 microscope using a 40× oil objective (EC Plan‐Neofluar, 1.3 NA), and eGFP‐Atx3 aggregate‐containing neurons were counted versus the number of cells presenting only diffuse eGFP‐Atx3 signal, as described previously.^[^ 59 ^]^ Aggregation was expressed as the percentage of cells containing aggregates.

Neuronal Cell Model—Image Quantification

Images were quantified using the image analysis software FIJI. Background intensity was subtracted from the user‐defined threshold and normalized to the cell body area. Synapses were evaluated using a semiautomatic macro designed for this purpose. The area of interest was randomly selected by using MAP2 staining for nontransfected cells and GFP staining for transfected cells. Dendritic length was measured in the area of interest, using one of the mentioned channels. To quantify the proteins of interest, images were subjected to a user‐defined intensity threshold to have defined protein clusters and user‐defined background intensity was subtracted in all images. Each protein cluster present in the area of interest was analyzed and intensity, area, and number measurements were obtained. Synaptic clusters were selected by overlapping PSD‐95 clusters of interest with thresholded VGluT1 clusters. All results were normalized to dendritic length. Image analysis was performed blinded to the experimental condition.

Neuronal Cell Model—Treatments

Cells (DIV 15) were treated with 10 µm CLR01 or CLR03 for 24 h before fixation.

C. elegans SCA3 Model—C. elegans Strains, Maintenance, and Synchronization

The wild‐type Bristol N2 strain was obtained from Caenorhabditis Genetics Center (University of Minnesota). The C. elegans model of SCA3, expressing full‐length human ATXN3 proteins with 130 glutamines [AT3q130: AM685 rmls263(P_F25B3.3_::AT3v1‐1q130::yfp)] throughout the animals’ nervous system was generated as described previously.^[^ 62 ^]^ Animals were grown at 20 °C on nematode growth medium agar plates seeded with E. coli OP50 strain, which was grown overnight at 37 °C and 180 rpm in Luria Broth (LB) media.^[^ 105 ^]^ To obtain an age‐synchronized population of eggs, gravid adults were treated with an alkaline hypochlorite solution (0.5 m NaOH, ≈2.6% NaClO) for 5 min. The eggs were washed in M9 buffer, resuspended in S‐medium to the appropriate egg number, and transferred into the 96‐well plates (drug assays). Batches of OP50 bacteria were grown overnight at 37 °C and 180 rpm in LB, pelleted by centrifugation, inactivated by three cycles of freeze/thawing, frozen at −80 °C, and then resuspended in S‐medium supplemented with cholesterol, streptomycin/penicillin, and nystatin (Sigma).

C. elegans SCA3 Model—CLR01 Preparation for C. elegans Assays

A 50 mm stock solution of CLR01 was prepared in dimethyl sulfoxide (DMSO). For each concentration to be tested, a stock solution was prepared at a 100× concentration in 100% v/v DMSO (D5879‐Sigma), from which work dilutions were prepared 2.4× concentrated. A final concentration of 1% v/v DMSO was used in all the conditions tested to avoid solvent‐specific developmental defects and toxicity.

C. elegans SCA3 Model—Drug Assay to Determine Compound Toxicity in C. elegans

The C. elegans Bristol strain N2 was used to determine the safe concentrations of CLR01. The assay was performed in a 96‐well plate format in liquid culture.^[^ 8 ^]^ Each well contained a final volume of 60 µL, comprising 20–25 animals in the egg stage, CLR01 at final concentrations 0.0001, 0.001, 0.01, 0.1, 1, 10, 50, or100 µm, and OP50 bacteria at a final OD at 595 nm of 0.6–0.7. Worms were grown with continuous shaking at 180 rpm at 20 °C (Shel Lab incubator shaker), and the bacteria OD was measured daily for seven days in the microplate reader (NanoQuant Plate‐Tecan). The effect of the compound on C. elegans physiology was monitored by the rate at which the E. coli food suspension was consumed as a readout for C. elegans growth, survival, or fecundity. DMSO 1% (drug vehicle) and DMSO 5% (toxic condition) were used as controls.

C. elegans SCA3 Model—Chronic and Postsymptomatic Treatment of C. elegans with CLR01

Chronic pre‐ and postsymptomatic C. elegans treatments with CLR01 were performed in 96‐well plates in liquid culture. Presymptomatic, chronic, treatment plates comprised: 45–50 animals per well in S‐medium, CLR01 at the respective concentrations in DMSO 1% [100–0.0001 µm], and OP50 bacteria (previously inactivated by cycles of freeze/thawing) resuspended in S‐medium complete to a final OD at 595 nm of 0.9. Treatment was initiated at the egg stage until day four posthatching. Postsymptomatic treatment with CLR01 was performed as described above except that animals were kept in drug vehicle (DMSO 1%) until day four posthatching. Starting on day 4, animals were washed daily and incubated with freshly prepared 0.1 µm CLR01 or 1% DMSO as a control. The plates were sealed to prevent evaporation and incubated at 20 °C with shaking at 180 rpm (Shell lab incubator shaker) for the indicated times.

C. elegans SCA3 Model—C. elegans Motility Assays

A method described previously^[^ 62 ^]^ was used to determine the percentage of animals with motor impairments after treatment with CLR01 at the indicated concentrations. Briefly, ≈10 animals were transferred simultaneously into the middle of a freshly seeded plate equilibrated at 20 °C. Animals remaining inside a 1 cm circle after 1 min were scored as locomotion‐defective. In the chronic presymptomatic treatment with CLR01, the motor phenotype was assessed on day four posthatching. A total of 200–250 animals were scored in 4–5 independent assays for each condition. In the postsymptomatic treatment with CLR01, animals' motor phenotype was assessed on day 4 (before treatment initiation), and the impact of CLR01 postsymptomatic treatment was evaluated on days 6 and 8. A total of ≈200 animals were scored in 4 independent assays for each condition. All assays were performed at room temperature (≈20 °C) using synchronized animals grown at the same temperature. All animals were scored at the same chronological age, and the scoring was performed blinded to treatment.

C. elegans SCA3 Model—C. elegans Aggregation Assays

In vivo images of the head region of AT3q130 animals were obtained using confocal microscopy (Olympus FV1000) under a 60× oil objective (NA = 1.35). The animals were transferred to 3% agarose pads and immobilized using 5 mm levamisole. Z‐series imaging was acquired for vehicle‐ and 0.1 µm CLR01‐treated animals using the 515 nm laser line for excitation of yellow fluorescent proteins. The pinhole was adjusted to 1.0 Airy unit of optical slice, and a scan was acquired every 0.5 µm along the Z‐axis. The obtained images were analyzed using a MeVisLab tool, as described previously.^[^ 62, 106 ^]^ Two parameters were measured: area of aggregates/total area and number of aggregates/total area. The values shown were the mean (normalized to vehicle‐treated control). Three independent trials were performed, and 28 and 30 animals were analyzed in the vehicle and CLR01 treatments.

Mouse SCA3 Model—Animal Housing Conditions

Female transgenic SCA3 and WT‐littermates (background C57BL/6J) exposed to CLR01/vehicle treatments were housed at weaning in groups of 5–6 animals in filter‐topped polysulfone cages 267 × 207 × 140 mm (370 cm^2^ floor area) (Tecniplast, Buguggiate, Italy), with corncob bedding (Scobis Due, Mucedola SRL, Settimo Milanese, Italy) in a conventional animal facility. The animals were kept under regular laboratory housing conditions: an artificial 12 h light/dark cycle (lights on from 8 am to 8 pm), with a relative humidity of 50–60% and an ambient temperature of 21 ± 1 °C. Mice were under a standard diet (4RF25 during the gestation and postnatal periods and 4RF21 after weaning, Mucedola SRL, Settimo Milanese, Italy) and water ad libitum. Sentinel animals in the same housing room were used to monitor the health status of the mice, according to FELASA guidelines, confirming the Specified Pathogens status. Humane endpoints for the experiment were defined as a 20% reduction of the body weight, inability to reach food and water, presence of wounds in the body, or dehydration. Still, they were not needed in practice as the ages tested in this study were within a range in which the animals did not reach these endpoints. Standardized environmental enrichment in the animal facilities (soft paper and creased paper for nesting behavior) was used.

Mouse SCA3 Model—CLR01 Treatment Groups

Animals were distributed among the different treatment groups in a randomized manner: WT vehicle (n = 19), WT CLR01 (n = 13), SCA3 vehicle (n = 20), and SCA3 CLR01 (n = 14). DNA extraction, animal genotyping, and CAG repeat size analyses were performed as described previously.^[^ 107 ^]^ The CAG repeat mean was not different among SCA3 groups (mean ± standard deviation (SD); [min–max]CAG; vehicle: 140 ± 2.56; [135–143]CAG; CLR01: 139 ± 3.04; [136–144]CAG). CLR01 (1 mg kg^−1^ per day) or vehicle (0.9% saline) was administered subcutaneously, daily, and 5 times per week from 4 to 22 weeks of age. CLR01 solution was prepared freshly every week in 0.9% saline. The injection volume was 100 µL and was administered by an independent experimenter (not involved in performing behavioral testing).

Behavioral analyses were performed during the diurnal period in groups of 5–6 animals per cage of CMVMJD135 hemizygous transgenic mice and WT littermates treated with CLR01 or vehicle.^[^ 6 ^]^ The same experimenter, blind to treatment, always performed the behavioral tests. At the end of the preclinical trial, the animals were euthanized according to their final purpose: either by decapitation or exsanguination perfusion with saline or PFA 4%. In the latter case, the animals were deeply anesthetized with a mixture of 150 mg kg^−1^ ketamine hydrochloride and 0.3 mg kg^−1^ medetomidine.

Mouse Phenotypic Analysis—Body Weight

All the animals were weighed at four weeks of age before treatment initiation, and their body weight was monitored every two weeks during the entire study.

Mouse Phenotypic Analysis—Beam Walk Balance Test

The test was performed as described previously.^[^ 108 ^]^ Animals were trained by the experimenter for three consecutive days using a 12 mm square‐shaped, 1 m long beam. Each animal had to traverse the beam 3 times during the learning phase. On day four, animals were tested in two different beams: the training 12 mm squared and an additional round beam of a 17 mm diameter. Animals were allowed to fail twice during the test, either by turning around or falling. Quantification of the time each animal took to traverse the beams was registered by the experimenter.

Mouse Phenotypic Analysis—Motor Swimming Test

To analyze voluntary locomotion, the mice were trained for two consecutive days, three trials for each animal, to traverse a clear, 100 cm long, water tank to a safe platform at the end. The latency to cross the water tank was measured from a distance of 60 cm. The water temperature was monitored and set at 23 °C using a thermostat.^[^ 108 ^]^

Mouse Phenotypic Analysis—Spontaneous Activity

The mice were transferred to a 15‐labeled‐square arena (55 × 33 × 18 cm). The number of squares traveled in the arena for 1 min was counted. The experimenters, blinded to treatment, registered the gait quality as normal or abnormal.

Mouse Phenotypic Analysis—Neuropathology Analysis

SCA3 and WT littermate mice were deeply anesthetized and transcardially perfused with PBS, followed by PFA 4% in PBS. Spinal cords were harvested and postfixed overnight in a fixative solution. The following day, spinal cords were transferred to a sucrose 30% solution and further sliced in a vibratome. 40 µm thick, free‐floating sections were stained with cresyl violet or processed for immunohistochemistry with mouse anti‐Atx3 (1:500, Millipore, 1H9, Cat #MAB5360, Lot #4080105). Atx3 positive inclusions, pyknotic cells, and motor neurons were quantified in the ventral horn (100% and 50% of coverage, respectively; n = 5–6 animals per group; 6 slices per animal) and normalized to the total area using an Olympus BX51 stereological microscope and a Visiopharma integrator system software. The integrated density of amyloid fibrils and fibrillar oligomers obtained by immunostaining with anti‐OC antibody was measured in mosaic pictures obtained using a 20× objective in an Olympus LPS Confocal FV3000 microscope. The sum of the pixel intensity values (integrated density) was obtained within the area drawn (gray matter of the spinal cord, using the polygon option) in the Fiji software. The integrated density was then normalized for each corresponding area. Negative controls (no primary antibody added) were performed in all immunostaining experiments. All the samples were codified immediately after euthanasia using numeric codes, and the quantification was performed blinded to genotype and treatment.

Statistical Analysis

Statistical analysis was performed using the IBM SPSS statistics 23 software (SPSS, RRID:SCR_002865) and GraphPad Prism 8.0.1 or 10 software.

Statistical Analysis—ThT Assays

The kinetic parameters v 50 or t 50 were compared across treatment groups using a mixed‐effects analysis restricted maximum likelihood (REML) for Atx3 13Q and a repeated measures one‐way ANOVA for Atx3 77Q (factor treatment, 3 levels). REML approach was chosen for Atx3 13Q analysis instead of repeated‐measures ANOVA because one biological replicate was missing data for CLR03 (1:10) incubation. This was followed by Dunnet's multiple comparisons test, which was applied to compare Atx3 incubation with CLR01 or CLR03 against the control. Data were reported as a mean ± standard error of the mean (SEM). Statistical analyses were performed in GraphPad Prism 10 (GraphPad Software, San Diego, CA). p < 0.05 was considered statistically significant.

Statistical Analysis—C. elegans

Continuous variables with normal distributions were analyzed according to Kolmogorov–Smirnov (K–S) test, p > 0.05, and with homogeneity of variance assessed by Levene's test, p > 0.05. When these two parameters were confirmed, groups were compared using one‐way ANOVA (factor: treatment) using the Dunnet test for post‐hoc analysis, and when the comparison was only between two groups, an independent‐sample t‐test was used. In the toxicity tests, in the comparison between groups, a nonlinear regression curve fit was used for sigmoidal curves analyzed using a least squares model with two parameters, IC_50_, and Hill Slope values. Results were presented as the mean ± SEM.

Statistical Analysis—Mice

The experimental unit in this study was a single animal. For the motor phenotype analysis, sample size calculations were performed for each behavioral test, assuming a power of 0.8 and a significance level (p‐value) of 0.05. The effect size was calculated to be 50%, using mean and standard deviation values obtained previously for transgenic and control groups for each test.^[^ 6, 8 ^]^ Variance homogeneity was evaluated using Levene's test. Continuous variables with normal distributions (K–S test, p > 0.05) were analyzed using a unidirectional or repeated‐measure ANOVA. Appropriate post‐hoc tests were used to follow up on any significant effects of genotype, treatment, or interaction effects found in the ANOVA having in consideration the unequal sample size of the experimental groups. Behavioral data were subjected to nonparametric Mann–Whitney U‐test or Kruskal–Wallis H‐tests when variables were noncontinuous or when continuous variables did not present a normal distribution (Shapiro–Wilk test, p < 0.05). Outliers were considered when the value obtained was 1.5 × IQR. A critical value for significance of p < 0.05 was used throughout the study.

Statistical reports of all the data analyzed can be found in Tables S2–S11 (Supporting Information).

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information