Introduction

Targeted protein degradation (TPD) is an emerging therapeutic strategy that leverages endogenous protein quality control systems to selectively degrade disease-associated proteins. TPD utilises bifunctional molecules in an induced proximity approach1, which brings together target proteins and key players in the ubiquitin proteasome (UPS) or the lysosomal autophagy degradation pathways, to degrade target proteins2. The induced proximity strategy has the potential to revolutionise the treatment of neurodegenerative diseases characterised by the accumulation of misfolded or toxic proteins, including Alzheimer’s disease, Parkinson’s disease and amyotrophic lateral sclerosis (ALS), by promoting the clearance of aberrant misfolded proteins.

There are currently over 200 ALS-associated mutations identified in the SOD1 gene3 that interfere with the folding of the encoded protein, destabilising the structure, and resulting in aggregation and the acquisition of a toxic function4,5. Several studies have also implicated misfolded forms of wild-type (WT) SOD16,7,8,9,10 in protein aggregation and gain-of-toxic function in ALS, although this remains controversial11,12. What is clear is that the misfolding and accumulation of toxic SOD1 species disrupts various cellular functions and is central to disease pathogenesis13,14,15, making misfolded SOD1 a compelling target for therapeutic strategies aimed at facilitating protein degradation.

A key challenge in targeting toxic SOD1 species for degradation is the need to preserve the natively folded WT species. SOD1 knockout mice exhibit early onset sarcopenia and increased incidence of hepatocarcinoma16, and reports of rapid disease progression and early death of patients with homozygous mutations for SOD1 have raised concern over the consequences of targeting WT SOD117,18,19. Similarly, targeting the WT forms of other neurodegenerative disease-associated proteins, including TDP-4320 and Huntingtin[21](https://www.nature.com/articles/s41467-025-65481-w#ref-CR21 “Kingwell, K. Double setback for ASO trials in Huntington disease. https://www-nature-com.ezproxy.uow.edu.au/articles/d41573-021-00088-6

(2021).“), has accelerated disease progression in both pre-clinical models and in human trials.

Given the need to differentially target misfolded, toxic forms of intracellular SOD1 for degradation, while sparing the natively folded form, TPD provides an attractive therapeutic strategy for ALS. Over the past decade, proteolysis targeting chimeras (PROTACs) have emerged as a promising strategy for the degradation of proteins involved in cancer onset and progression22,23,24,25,26. PROTACs are heterobifunctional molecules consisting of two ligands joined by a flexible linker—one ligand binds the target protein and the other recruits an E3 ubiquitin ligase. Bringing these two molecules together in close proximity facilitates the ubiquitination of the target protein and its subsequent degradation by the ubiquitin proteasome system (UPS)2. While conventional PROTACs utilise small-molecule ligands to bind target proteins, naturally existing protein binding partners or antibodies specific for the target protein, termed BioPROTACs, have recently been successfully developed27,28,29.

In this work, we design and test a series of BioPROTACs specifically targeting misfolded forms of SOD1 for degradation. We explore various chimeras of different SOD1 binding proteins and E3 ligases using a panel of seven single-chain variable fragments (scFvs) derived from monoclonal antibodies that bind to aggregated SOD1 in SOD1-ALS patient tissue8, soluble aggregated SOD130, and toxic seeding species of mutant and WT misfolded SOD18,31,32. Each of these misfolded SOD1 scFvs is fused with a panel of eight proteins possessing the ubiquitination functionality of E3 ligases. Using in vitro assays across three cell lines to screen these panels, we identify a lead BioPROTAC candidate we term MisfoldUbL, as it is specific for misfolded SOD1 and functions via the ubiquitin ligase pathway. In a compound transgenic MisfoldUbL/SOD1G93A mouse line, we show that expression of MisfoldUbL delays disease progression, reduces the amount of insoluble SOD1 in the brain, and protects against spinal cord motor neuron loss and denervation at neuromuscular junctions. Together, our results demonstrate that a BioPROTAC can specifically reduce misfolded protein species, leading to disease-modifying effects, supporting this approach in therapeutic applications.

Results

BioPROTAC design

ProMIS™ Neurosciences Inc. (Toronto, Canada) had previously generated a panel of seven monoclonal antibodies raised against the electrostatic loop of SOD1 (Fig. 1A), a region of SOD1 that is inaccessible when the protein is natively folded33. These antibodies have been validated as specifically binding misfolded forms of SOD114,34. To condense their size for use in a BioPROTAC, we generated single-chain variable fragment (scFv) sequences from these seven antibody clones, comprising VH and VL regions joined by a (G4S)3 linker (Fig. 1A). For the initial scFv screen, the C-terminal catalytic domain of Hsc70-interacting protein (CHIP) was truncated to remove the natural binding domain (CHIP∆TPR), and was then fused to the misfolded SOD1 scFv sequences via a GSGSG linker (Fig. 1B). This resulted in the generation of seven unique chimeric BioPROTACs, referred to herein as BPs 1–7. An scFv to an unrelated protein (β-galactosidase) fused to CHIP∆TPR was used as a control in all in vitro assays. The BioPROTACs were designed to bind misfolded SOD1 via their scFv domain, resulting in ubiquitination by the E3 ligase domain and subsequent proteasomal degradation of misfolded SOD1. The effectiveness of degradation could then be measured by the depletion in EGFP fluorescence (Fig. 1C).

Fig. 1: Design and proposed mechanistic action of the SOD1-targeting BioPROTACs.

A Antibodies were raised against the SOD1 electrostatic loop (magenta) and single-chain variable fragments (scFvs) were generated from the antibody sequences, with heavy (VH) and light (VL) chains joined by a flexible (G4S)3 motif. Created in BioRender. Chisholm, C. (2025) https://BioRender.com/mg3bqw4. B For the initial scFv screen, the E3 ligase CHIP was truncated to remove the binding domain, and the catalytic domain was then fused to misfolded SOD1 scFvs. FLAG and 6x His tags were added for detection. C BioPROTACs are proposed to bind misfolded SOD1 (mSOD1) via their scFv domain, resulting in ubiquitination by the E3 ligase domain and subsequent proteasomal degradation of misfolded SOD1, which can be measured through EGFP-mediated fluorescence. Created in BioRender. Chisholm, C. (2025) https://BioRender.com/w5c64g2.

BioPROTAC scFvs specifically engage misfolded SOD1 in diverse cell types

We used a panel of in vitro assays to assess the efficacy of BPs 1–7 in reducing cellular levels of aggregation-prone SOD1. We have previously shown that the expression of C-terminally EGFP-tagged SOD1 variants can lead to inclusion formation in cultured cells35,36. To this end, HEK293 cells were co-transfected to express either SOD1WT-EGFP or the highly aggregation-prone SOD1A4V-EGFP with each of the BioPROTACs, and the level of SOD1-EGFP was quantified. Three of the BioPROTACs (BP1, BP4 and BP5) slightly reduced SOD1WT-EGFP levels, ranging from 7 to 8% reduction (Fig. 2A). All BioPROTACs except BP7 reduced the amount of misfolded SOD1A4V-EGFP in cells compared to the control, ranging from 17 to 38% reduction (Fig. 2B). This reduction was confirmed with immunoblotting (Fig. 2C). The negligible changes in SOD1WT-EGFP level compared to the significant changes observed for SOD1A4V-EGFP levels strongly indicate specificity of the BioPROTACs towards misfolded forms of SOD1, a finding supported by the unchanged levels of endogenous SOD1 in the presence of the BioPROTACs (Supplementary Fig. 1).

Fig. 2: BioPROTAC scFvs engage misfolded SOD1, leading to a decrease in SOD1 aggregation.

Reduction of A SOD1WT-EGFP and B SOD1A4V-EGFP fluorescence in HEK293 cells expressing BioPROTACs relative to cells co-transfected with the control over 48 h. C Immunoblot of lysates from cells transfected with SOD1A4V-EGFP and the BioPROTAC panel. D Schematic representation of saponin permeability assay for measuring proportion of cells with aggregates. Created in BioRender. Chisholm, C. (2025) https://BioRender.com/vx24bql. E The number of cells containing insoluble SOD1A4V-EGFP aggregates in HEK293 cells expressing BioPROTACs relative to cells co-transfected with the control was quantified using the saponin permeability assay. F Immunocytochemistry was used to assess BioPROTAC and SOD1A4V-EGFP expression in HEK293 cells. G The reduction in SOD1A4V-EGFP fluorescence compared to control was determined when expression levels of the BioPROTACs were normalised. H A weighted heat map comparing the relative efficacy of BioPROTACs in the various assays, with BP2 identified as the most effective BioPROTAC. I The reduction in the number of cells containing insoluble SOD1-EGFP aggregates following expression of BP2 across a range of SOD1 variants relative to cells co-transfected with the control. J Cell counts for HEK293 cells untransfected or transfected with BP2 for 24 or 48 h. K Co-immunoprecipitation was used to assess the specificity of BP2 binding for misfolded SOD1 over the WT form. For all graphs, bars represent mean ± SEM (* P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001). Statistical significance was determined using (A, B) two-way ANOVA paired with Dunnett’s multiple comparisons test, (E, I) ordinary one-way ANOVA paired with Dunnett’s multiple comparisons, (F, G) Kruskal–Wallis one-way ANOVA paired with Dunn’s multiple comparisons test or (J, K) unpaired Student’s t-tests. Blots are representative from at least 3 independent experiments. Raw data, complete western blots, total protein images and exact P-values are shown in Source data file and Supplementary Table 2.

Previous work has shown that expression of SOD1 variants in cultured cells leads to the formation of large insoluble aggregates that are reminiscent of the inclusions observed in motor neurons in tissues from ALS patients37. Therefore, we investigated the ability of our BioPROTACs to reduce the formation of insoluble aggregates in cells (Fig. 2D). There was a significant reduction in the number of cells with insoluble SOD1A4V-EGFP aggregates upon co-expression of BP2 (75 ± 4%), BP3 (76 ± 3%), BP4 (55 ± 5%) and BP6 (73 ± 4%) compared to the control (Fig. 2E), demonstrating the efficacy of these BioPROTACs at reducing toxic SOD1 accumulation. Similar effects were also observed in two neuron-like cell lines; Neuro-2A and SH-SY5Y, with significant reductions in expression levels of SOD1A4V-EGFP and SOD1G93A-EGFP, and a reduction of insoluble aggregates by one or more BioPROTACs in each assay, including BP2 in all assays (Supplementary Fig. 2).

To investigate the relationship between BioPROTAC expression levels and the reductions previously observed in misfolded SOD1, we analysed cells by immunocytochemistry (Supplementary Fig. 1F). There were significant increases in expression levels of BP1, BP3, BP4, BP5 and BP7 compared to control (Fig. 2F). After normalising for the level of expression, all the BioPROTACs were found to reduce SOD1A4V-EGFP levels compared to the control, with BP2 and BP3 having the greatest effect, reducing levels by 78 ± 0.1% and 93 ± 0.1% respectively (Fig. 2G). Comparing the efficacy of BioPROTACs across the assays performed demonstrated that BP2 was the most effective at reducing both misfolded SOD1 levels and aggregation in cells (Fig. 2H).

We next investigated if BP2 could reduce the cellular levels and aggregation of other SOD1 mutants. With over 200 pathogenic mutations identified19, a common unfolded state exposing the scFv epitope would be beneficial for a broad-based SOD1-directed BioPROTAC to have therapeutic relevance. Using the cell-based aggregation assay described above, we found that expression of BP2 significantly reduced the proportion of cells with insoluble aggregates compared to control across the range of mutant SOD1 variants tested (A4V, G93A, G85R, D90A, V148G, H46R, G37R, C6G and E100G) (Fig. 2I). Thus, BP2 has broad specificity for misfolded SOD1 variants. To investigate cell toxicity, we compared cell counts between untransfected cells and cells transfected with BP2 over 24 and 48 h, and found no difference (Fig. 2J). To confirm that BP2 binds preferentially to misfolded forms of SOD1, a co-immunoprecipitation assay was employed. As expected, BP2 bound SOD1A4V-EGFP more effectively (7-fold) than SOD1WT-EGFP (Fig. 2K), supporting its specificity for the misfolded form of the protein. Hence, BP2 was selected as the lead candidate to progress for further BioPROTAC optimisation, and used in subsequent assays and experiments described below.

BioPROTAC-mediated reduction of mutant SOD1 is E3 ligase-dependent

The UPS-mediated mechanism of protein degradation harnessed by both PROTAC and BioPROTAC approaches requires ubiquitination of the target protein by an E3 ligase. To optimise the degradation of misfolded SOD1 observed in the initial scFv screen, we tested a panel of eight E3 ligases from across the four family groups (U-BOX, HECT, RING and Ring between Ring). Selection of candidates was based on strong expression levels in relevant cell lines and tissue types, and retention of ubiquitination activity when truncated38,39,40,41,42,43,44,45. The catalytic domains of selected E3 ligases, UBE4A, UBOX5, NEDD4L, UBR5, RNF4, βTrCP and Parkin (with intrinsic substrate binding domains removed), were fused to BP2, the most effective scFv from the initial BioPROTAC screen. We configured the scFv and detection tags of each chimera to preserve the relative position of the catalytic domain relative to the rest of the protein in the native E3 ligase (Fig. 3A).

Fig. 3: BioPROTAC efficacy is dependent on the E3 ligase component.

A A panel of eight E3 ligases were fused to the most effective scFv determined previously. The panel included representatives from the 4 families of ligases. Positioning of scFv and detection tags were dependent on position of the catalytic domain in the native E3 ligase. Levels of the BioPROTACs were assessed in the B soluble and C insoluble fractions from HEK293 lysates. D Immunocytochemistry was used to assess BioPROTAC and SOD1G93A-EGFP expression in HEK293 cells. E The reduction in SOD1G93A-EGFP fluorescence compared to control was determined when expression levels of the BioPROTACs were normalised. F The number of cells containing insoluble SOD1G93A-EGFP aggregates in HEK293 cells expressing BioPROTACs relative to cells co-transfected with the control was quantified using the saponin permeability assay. G Immunoblot of lysates from cells transfected with SOD1G93A-EGFP and the ligase panel. For all graphs, bars represent mean ± SEM (** P < 0.01, **** P < 0.0001). Statistical significance was determined using (B, C) ordinary one-way ANOVA paired with Tukey’s multiple comparisons test, (D, E) Kruskal–Wallis one-way ANOVA paired with Dunn’s multiple comparisons test or (F, G) ordinary one-way ANOVA paired with Dunnett’s multiple comparisons test. Blots are representative from at least 3 independent experiments. Raw data, complete western blots, total protein images and exact P-values are shown in Source data file and Supplementary Table 2.

BioPROTAC candidates containing RNF4, UBE4A, Parkin, and CHIP catalytic domains showed variable soluble expression in HEK293 cells, with UBE4A and CHIP constructs displaying the highest expression (Fig. 3B). BioPROTACs containing βTrCP, UBOX5, NEDD4L and UBR5 catalytic domains were only detected in the insoluble fraction (Fig. 3C). Interestingly, all soluble BioPROTAC constructs had the scFv in an N-terminal configuration relative to the ligase domain. Immunostaining confirmed that soluble BioPROTAC ligases were highly expressed, with significantly lower expression observed with insoluble ligases (Fig. 3D). These soluble and relatively highly expressed BioPROTACs (i.e. CHIP, UBE4A, Parkin and RNF4) significantly reduced levels of misfolded SOD1G93A-EGFP in HEK293 cells compared to control (Fig. 3E). Moreover, the BioPROTACs containing CHIP, UBE4A and Parkin significantly reduced the number of cells with insoluble SOD1G93A-EGFP aggregates (Fig. 3F). CHIP and Parkin also showed significant reduction of SOD1G93A-EGFP in HEK293 cells compared to control by immunoblot (Fig. 3G). Similar effects were observed using an alternate SOD1 mutant, SOD1A4V (Supplementary Fig. 3A–C), and in a second cell line, SH-SY5Y (Supplementary Fig. 3D, E), with BioPROTACs containing CHIP, UBE4A and Parkin reducing levels of misfolded SOD1 compared to control. CHIP-, UBE4A- and Parkin-containing BioPROTACS displayed comparable capacity to reduce levels and aggregation of misfolded SOD1 across these assays (Supplementary Fig. 3F). Considering these data, and the high expression levels of CHIP in spinal cord tissue44,45, the BioPROTAC containing BP2 and CHIP catalytic domain was selected as the lead candidate for further testing.

BioPROTACs function via both lysosomal and proteasomal degradation pathways

To investigate the E3 ligase dependency of BioPROTAC-mediated degradation of misfolded SOD1, we compared the efficacy of the BP2-CHIP chimera with a construct containing only the scFv component of BP2 (BP2ΔCHIP). HEK293 cells were co-transfected to express SOD1G93A-EGFP and either control, BP2 or BP2ΔCHIP. The active BP2 chimera reduced the proportion of cells containing SOD1G93A-EGFP aggregates compared to control (45 ± 10%), while BP2ΔCHIP had no effect, indicating the importance of E3 ligase activity for degradation of misfolded SOD1 (Fig. 4A). Interestingly, while the BP2 chimera reduced the number of cells containing aggregates of the more aggregation-prone mutant SOD1A4V-EGFP (58 ± 0.1% reduction compared to control), some low-level activity was still observed in the absence of the catalytic domain, with BP2ΔCHIP reducing aggregation of SOD1A4V-EGFP, by 20 ± 3% (Fig. 4A).

Fig. 4: BioPROTACs degrade misfolded SOD1 via both lysosomal and proteasomal degradation pathways.

A The number of cells containing insoluble SOD1-EGFP aggregates in HEK293 cells expressing either BP2 or BP2ΔCHIP and SOD1G93A-EGFP or SOD1A4V-EGFP relative to cells co-transfected with the control. B Levels of SOD1 and Hsp70 in cells transfected with increasing amounts of BP2 or BP2ΔCHIP. C The number of cells with insoluble SOD1A4V-EGFP aggregates after increasing concentrations of MG132 was quantified for BP2 and BP2ΔCHIP. D The number of cells with insoluble SOD1A4V-EGFP aggregates after increasing concentrations of Baf. A1 was quantified for BP2 and BP2ΔCHIP. E Levels of SOD1 in cells transfected with BP2 or BP2ΔCHIP then treated with vehicle, MG132 or Baf.A1. For all graphs, bars represent mean ± SEM (* P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001). Statistical significance was determined using (A) repeated measures or (B–D) ordinary one-way ANOVA paired with Tukey’s multiple comparisons test. Blots are representative from at least 3 independent experiments. Raw data, complete western blots images and exact P-values are shown in Source data file.

The effect of BP2 on misfolded SOD1 degradation was concentration-dependent. HEK293 cells were co-transfected to express SOD1A4V-EGFP with increasing levels of either BP2 or BP2ΔCHIP, and lysates were collected for immunoblotting. The presence of the ligase in BP2 resulted in greater reductions in cellular SOD1G93A-EGFP levels compared to BP2ΔCHIP at all levels tested. Only the highest amount of BP2ΔCHIP reduced SOD1G93A-EGFP levels compared to control (Fig. 4B). To investigate whether the presence of truncated CHIP in the BioPROTAC affected its endogenous function, we assessed the levels of a known CHIP substrate Hsp70 (HSPA1A) in cells expressing BP2 and BP2ΔCHIP, and found no difference (Fig. 4B). These data suggest that endogenous substrates of CHIP are not targeted by BP2. Immunoblotting and an SOD1 activity assay confirmed the specificity of BP2 for the misfolded form of SOD1 over the WT form (Supplementary Fig. 4).

It has been proposed that scFvs that target misfolded SOD1 function by binding and preventing aggregation without reducing the overall level of misfolded SOD1, rather than degrading it46. However, a recent study using an scFv against α-synuclein showed that an scFv-α-synuclein complex was processed through the lysosomal degradation pathway47. To investigate the involvement of the two major protein degradation pathways in BP2 and BP2ΔCHIP action, we used a pharmacological approach. Treatment with proteasome inhibitor MG132 resulted in increased aggregate formation when cells expressed BP2 but not BP2ΔCHIP (Fig. 4C), suggesting the UPS plays a major role in the degradation mechanism when the E3 ligase domain is present. Treatment with Bafilomycin A (Baf.A1), a small molecule that inhibits autophagosome-lysosome fusion and acidification48, resulted in a partial blocking of effect for cells expressing BP2, but complete loss of effect for cells expressing BP2ΔCHIP (Fig. 4D). Similarly, SOD1A4V clearance was abolished when HEK293 cells transfected with BP2 were exposed to MG132 but not when treated with Baf.A1 (Fig. 4E). Taken together, these results suggest that BP2ΔCHIP decreases SOD1 levels via lysosome-mediated degradation, whilst BP2 harnesses both the lysosomal and UPS pathways to degrade SOD1, leading to improved efficacy of BP2.

Expression of MisfoldUbL in neurons prevents weight loss and slows disease progression in SOD1G93A mice

Having established that a BioPROTAC composed of scFv2 and CHIP (BP2) was the most effective at degrading misfolded SOD1 with minimal off-target effects, we moved to investigate in vivo efficacy. Henceforth, we refer to BP2 as MisfoldUbL, to reflect the targeting of misfolded SOD1 via the ubiquitin ligase pathway. We generated a transgenic mouse employing the human synapsin 1 promoter to restrict expression to neurons only49,50,[51](https://www.nature.com/articles/s41467-025-65481-w#ref-CR51 “Savell, K. E. et al. A Neuron-Optimized CRISPR/dCas9 Activation System for Robust and Specific Gene Regulation. eNeuro 6, https://doi.org/10.1523/eneuro.0495-18.2019

(2019).“). Immunohistochemistry and immunoblotting showed MisfoldUbL expression in both brain and spinal cord, but not liver (Fig. 5A, B). There was a 26-fold higher MisfoldUbL level in the brain than spinal cord (Fig. 5C).

Fig. 5: MisfoldUbL is expressed in transgenic mice and affects phenotypic characteristics in SOD1G93A mice.

A Immunohistochemistry to assess expression of MisfoldUbL in neuronal cells in brain tissue of WT/mUbL mice. Images are representative of n = 6 mice. Scale bar represents 50 µm and MisfoldUbL is indicated with black arrows. B Immunoblot to assess expression of MisfoldUbL in brain, spinal cord and liver in WT/mUbL mice. C Relative levels of MisfoldUbL in brain versus spinal cord. D Schematic representation of experimental setup. Heterozygous SOD1G93A mice and heterozygous MisfoldUbL mice were crossed to obtain four experimental groups; non-transgenic WT/WT, WT/MisfoldUbL, SOD1G93A/WT and SOD1G93A/MisfoldUbL mice. Created in BioRender. Chisholm, C. (2025) https://BioRender.com/zjatewp. E–L Phenotypic data comparing SOD1G93A/WT (n = 11 male and n = 10 female) and SOD1G93A/MisfoldUbL mice (n = 11 male and n = 12 female) are shown. Data of control groups WT/WT (n = 24) and WT/MisfoldUbL (n = 24) are shown in Supplementary Fig. 5. Male and female mice were assessed for E, F weight gain, G, H motor function via latency to fall on the rotarod, I, J disease progression via neurological score and K, L survival. Results represent mean ± SEM (shading) (* P < 0.05, ** P < 0.01, **** P < 0.0001). Statistical significance was determined using (C, I inset and J inset) unpaired, two-tailed t-tests, (E–J) repeated measures one-way ANOVA paired with Tukey’s multiple comparisons test or (K, L) log-rank Mantel-Cox test. Blots are representative from at least 3 independent experiments. Raw data, complete western blots, total protein images and exact P-values are shown in Source data file.

To investigate the effect of MisfoldUbL expression on the ALS phenotype observed in SOD1G93A mice, we crossbred MisfoldUbL mice to generate SOD1G93A/MisfoldUbL mice (Fig. 5D). SOD1G93A mice have previously shown sex-specific differences in disease phenotype52,53,54, and we therefore assessed for sex-specific phenotypic changes. Male SOD1G93A/MisfoldUbL mice displayed a significant increase in body weight compared to SOD1G93A/WT mice (5.5 ± 0.16%) (Fig. 5E). No significant difference in body weight was observed between SOD1G93A/MisfoldUbL and SOD1G93A/WT female mice (Fig. 5F). Both male and female WT/MisfoldUbL mice showed a small but significant decrease in body weight compared to WT/WT mice (1.0 ± 0.07% males, 1.3 ± 0.08% females) (Supplementary Fig. 5A, B).

In the SOD1G93A mouse model, motor function is a primary measure for disease progression and severity, and was assessed weekly using an accelerating rotarod. SOD1G93A/MisfoldUbL male mice exhibited a longer latency to fall than SOD1G93A/WT mice (Fig. 5G). In contrast, female SOD1G93A/MisfoldUbL showed a shorter latency to fall than SOD1G93A/WT mice (Fig. 5H). For the control groups, as expected, both male and female WT/MisfoldUbL mice showed no significant difference in rotarod performance compared to WT/WT mice (Supplementary Fig. 5C, D).

While male SOD1G93A/MisfoldUbL mice showed no significant difference in disease onset compared to SOD1G93A/WT mice, longitudinal analysis of neurological score showed that SOD1G93A/MisfoldUbL mice exhibited a significantly slower disease progression (Fig. 5I). Female SOD1G93A/MisfoldUbL mice showed a delay in disease onset and a slower disease progression compared to SOD1G93A/WT mice (Fig. 5J).

Kaplan-Meier analysis showed no difference in survival between SOD1G93A/MisfoldUbL and SOD1G93A/WT for either male or female mice (Fig. 5K, L). However, the similarity in survival time masked an interesting observation that became strikingly evident as the study progressed. Of the male SOD1G93A/MisfoldUbL mice, 43% were euthanised at a defined ethical endpoint (based on weight loss) with a neurological score of 1 or 2. Even as these mice reached their weight endpoint, they remained extremely mobile and continued to display behaviours that required considerable limb strength, such as rearing and hanging from the wire food hopper. SOD1G93A/MisfoldUbL mice were also inquisitive and active in the cage (see Supplementary Videos 1 and 2 for video of SOD1G93A/MisfoldUbL and SOD1G93A/WT, respectively). No male SOD1G93A/MisfoldUbL mouse reached complete paralysis (neurological score of 4). A contingency analysis using two-sided Fisher’s exact test was performed to examine the relationship between genotype and the underlying reason for endpoint (i.e. paralysis or weight loss). The difference between these variables was significant for males, with SOD1G93A/WT mice more likely to reach paralysis (ALS 4) than SOD1G93A/MisfoldUbL mice. A similar trend was observed with female mice, where 83% of SOD1G93A/MisfoldUbL mice were euthanised for weight loss compared to 50% of the SOD1G93A/WT mice. Of the female SOD1G93A/MisfoldUbL mice that were euthanised for reaching the weight endpoint, 33% had ALS scores of 1 or 2; however, these differences did not reach statistical significance.

In summary, these results indicate that expression of the MisfoldUbL transgene in SOD1G93A mice delays disease progression and protects against progression of ALS symptoms, and this effect is more marked in male than female mice. Over time, the SOD1G93A/MisfoldUbL mice lost comparable body weight to SOD1G93A mice, resulting in no difference in survival data. However, expression of MisfoldUbL preserved motor control and function until the end stage of disease.

Expression of MisfoldUbL in brain and spinal cord neurons results in a reduction of insoluble SOD1, protection of motor neurons and a preservation of neuromuscular junctions

To investigate underlying ALS biochemical and physiological parameters associated with the observed phenotypic effects of MisfoldUbL expression at an early-symptomatic stage, a group of age-matched mice were sacrificed at 90 days old. Immunoblot analysis of soluble and insoluble fractions from brain and lumbar spinal cord homogenates showed a decrease in insoluble SOD1 in brain tissue at both early-symptomatic stage (Fig. 6A) and end stage (Fig. 6B) in SOD1G93A/MisfoldUbL mice compared to SOD1G93A/WT mice. However, this reduction in insoluble SOD1 was not observed in the spinal cord in either cohort, despite expression of the MisfoldUbL across all regions of the spinal cord (Supplementary Fig. 5E). There was no difference in soluble SOD1 in either cohort in the brain or spinal cord.

Fig. 6: Expression of MisfoldUbL in brain and spinal cord neurons results in a reduction of insoluble SOD1, protection of motor neurons and a preservation of neuromuscular junctions in SOD1G93A mice.

Immunoblots to assess levels of soluble and insoluble SOD1 protein in the brains and spinal cords of A early-symptomatic and B end-stage SOD1G93A/MisfoldUbL and SOD1G93A/WT mice. C Motor neuron numbers were counted in the ventral lumbar spinal cord of early-symptomatic and end-stage mice. D The number of fully innervated, partially innervated and denervated neuromuscular junctions of SOD1G93A/WT and SOD1G93A/MisfoldUbL mice. A representative image of a stained longitudinal gastrocnemius section is included showing the pre- (purple) and post-synaptic (green) markers. Scale bar represents 100 µm. A fully innervated NMJ is indicated in a solid black box, a partially innervated NMJ in a dashed yellow box, and a denervated NMJ in a dotted red box. Results represent mean ± SEM, n = 6 (early-symptomatic cohort) or 12 (end-stage cohort) mice per genotype (* P < 0.05). A–D Statistical significance was determined using unpaired two-tailed t-tests. Raw data, complete western blots images and exact P-values are shown in Source data file.

To investigate the retention of muscle function observed at end stage in the SOD1G93A/MisfoldUbL mice, we performed a motor neuron count in the lumbar spinal cord (Fig. 6C) and measured neuromuscular junction innervation in gastrocnemius muscles (Fig. 6D), and found the MisfoldUbL transgene conferred protection of both integral components of motor function. SOD1G93A/MisfoldUbL mice exhibited a 30 ± 6% increase in motor neuron number compared to SOD1G93A/WT mice at early-symptomatic stage, and 50 ± 15% at end stage. For the WT/MisfoldUbL and WT/WT control groups, there was no significant difference in motor neuron number at either the early-symptomatic stage or when the mice had reached 200 days old (Supplementary Fig. 5F). SOD1G93A/MisfoldUbL mice also displayed an increase in the number of fully innervated neuromuscular junctions in their gastrocnemius muscles at end stage. These findings are consistent with the functional observations noted for these mice, and further support the neuroprotective effect conferred by expression of MisfoldUbL.

Discussion

The aberrant aggregation of proteins is implicated in the onset and pathogenesis of a wide range of neurodegenerative disorders, including ALS. Increasing evidence indicates that misfolded protein oligomers produced as intermediates in the aggregation process are potent neurotoxic agents in disease[55](https://www.nature.com/articles/s41467-025-65481-w#ref-CR55 “Rinauro, D. J., Chiti, F., Vendruscolo, M. & Limbocker, R. Misfolded protein o