Novel Small-Molecule Analogues of IU1 Ameliorate Amyloid-β Mediated Toxicity in Alzheimer’s Disease Cell and Worm Models

TL;DR

Scientists designed new IU1 compounds that reduce amyloid-β toxicity in Alzheimer’s disease models, showing better results than the original IU1.

Contribution

New IU1 analogues (AA10 and AA51) were developed and shown to be more effective than IU1 in reducing Alzheimer’s-related toxicity.

Findings

IU1 and analogues reduced amyloid-β toxicity in cell and worm models.

Analogues AA10 and AA51 enhanced autophagy and proteasome activity more than IU1.

AA10 and AA51 rescued neuron loss and behavioral impairments in worms.

Abstract

Dysregulation of the deubiquitinating enzyme Ubiquitin-specific peptidase 14 (USP14) is implicated in several neurodegenerative diseases, and IU1, an allosteric inhibitor, has shown neuroprotective effects by reducing protein aggregate toxicity. This study aimed to develop new IU1 analogues and evaluate their ability to mitigate amyloid-β (Aβ) accumulation and toxicity in Alzheimer’s disease (AD) cell and Caenorhabditis elegans worm models. IU1 and 71 newly designed analogues identified using the AtomNet® virtual screening platform were assessed in an amyloid precursor protein-C terminal fragment/amyloid-β (APP-C99/Aβ)-producing AD cell model using a high-throughput toxicity assay. Lead compounds were further evaluated for their effects on neurodegeneration, behaviour, and survival. IU1 reduced Aβ-mediated toxicity and neurodegeneration in cell and worm models. Of the 71 analogues predicted to bind ubiquitin-specific peptidase 14 (USP14), two compounds, AA10 and AA51, showed >50% rescue of Aβ-induced toxicity and robust enhancement of autophagy and proteasome activity. In Caenorhabditis elegans, both compounds alleviated glutamatergic neuron loss and rescued behavioural impairments. IU1 and analogues exhibit protective effects against Aβ toxicity in AD models. Analogues AA10 and AA51 showed greater potency than IU1 and effectively enhanced proteostasis pathways. These findings support USP14 as a promising therapeutic target and provide a basis for the development of improved IU1-derived compounds for AD and related disorders.

Figures

Full text

1. Introduction

The burden of fatalities and impairment caused by neurological disorders is rapidly being identified as a global public health concern, and this burden is likely to rise within the next few decades. In the past 30 years, the absolute numbers of deaths have climbed by 39% and disability-adjusted life-years (DALYs; the sum of years of life lost and years lived with impairment) have increased by 15%, despite declines in infectious neurological disorders. A recent study estimates about 4.9 billion cases of brain disorders by 2050, representing a 22% increase from 2021 estimates [1,2].

Dysfunction of the ubiquitin–proteasome system (UPS) plays a central role in the accumulation of amyloid-β (Aβ) and other misfolded proteins in Alzheimer’s disease (AD) and several neurodegenerative disorders [3,4]. Our previous work demonstrated parallel impairments in both the autophagy–lysosomal pathway (ALP) and UPS in AD post-mortem brains and in a tetracycline-regulated MC65 cell model that produces amyloid precursor protein-C terminal fragment (APP-C99) and Aβ [5,6,7]. Ubiquitin-specific peptidase 14 (USP14), a proteasome-associated deubiquitinase (DUB), is a key negative regulator of proteasomal degradation. Its overactivation delays substrate turnover and exacerbates proteotoxic stress [8]. Because DUBs harbour well-defined catalytic pockets, they are considered druggable regulators of UPS and ALP function [5,9]. Previous reports show that USP14 inhibition confers neuroprotection in several neurodegenerative disease models, including tauopathy and prion disorders [10,11,12]. IU1, a small-molecule allosteric inhibitor of USP14, has been shown to attenuate the toxicity of multiple aggregation-prone proteins such as tau, TAR DNA-binding protein 43 (TDP-43), ataxin-3, and glial fibrillary acidic protein (GFAP) [13,14], highlighting USP14 as an attractive therapeutic target for reducing protein aggregation in AD.

In this study, IU1 was used as the reference compound for in vitro and in vivo assays. Consistent with prior findings, IU1 enhanced autophagy, promoted Aβ clearance, and reduced Aβ-induced toxicity in MC65 cells [7,15,16,17], and ameliorated neurodegeneration and behavioural deficits in a C. elegans AD model [18]. To identify more potent USP14 inhibitors, we applied the AtomNet^®^ deep-learning virtual screening platform (Atomwise, San Francisco, CA, USA), which yielded 71 novel IU1-derived ligands [19,20,21]. These compounds were screened in MC65 cells, and candidates achieving >50% rescue of Aβ-induced toxicity were prioritised for mechanistic studies and in vivo evaluation in Caenorhabditis elegans (C. elegans). Selected analogues enhanced Aβ clearance, improved proteostasis, and restored behavioural function in AD models. Overall, this work shows the neuroprotective activity of IU1 and identifies novel potent IU1 analogues capable of mitigating Aβ-associated toxicity in AD.

2. Results

2.1. IU1 Mitigates Aβ-Mediated Toxicity in Cell and Worm Models of AD

USP14 is a proteasome-associated DUB that negatively regulates proteasomal degradation, and its overactivation has been implicated in the accumulation of misfolded proteins and the progression of neurodegenerative diseases, including AD. IU1, a selective small-molecule inhibitor of USP14, has been previously shown to enhance proteasome activity and promote clearance of aggregate-prone proteins, such as tau, TDP-43, ataxin-3, and GFAP [13,14]. In the present study, we first assessed the efficacy of IU1 in our established human central nervous system (CNS)-derived neuroblastoma APP-C99/Aβ-producing AD cell model (MC65) to determine whether IU1 could mitigate Aβ-mediated toxicity and restore cellular proteostasis.

MC65 cells were cultured in the absence of tetracycline (−Tet) to induce APP-C99/Aβ expression. As expected, −Tet cells exhibited pronounced cytotoxicity, with only ~20% cell viability after four days of induction compared to +Tet control cells, which maintained near-normal viability (Figure 1A). Treatment with IU1 substantially rescued cell viability by ~70% improvement over untreated −Tet cells (Figure 1A).

In addition to its proteasome-enhancing properties, IU1 has been reported to modulate the ALP. To evaluate autophagic activity, we performed immunofluorescence staining for LC3 and p62, markers of autophagosome formation and cargo recognition, respectively. Untreated −Tet cells showed significant co-accumulation of LC3 and p62, indicative of impaired autophagic flux (Figure 1B). IU1 treatment markedly reduced this colocalisation, restoring levels to near those observed in +Tet controls. We also evaluated two previously reported analogues, IU1-47 and IU1-248, in the MC65 model [22]. Both analogues showed significant cytoprotective effects, with IU1-248 increasing cell viability by 60% and IU1-47 by 40% relative to untreated −Tet cells, validating the neuroprotective properties of the IU1 scaffold and its derivatives (Figure S1).

To validate IU1’s neuroprotective effects in vivo, we employed the C. elegans AD model, specifically the UA198 strain expressing human Aβ in glutamatergic tail neurons under the eat-4 promoter [18]. C. elegans possess five easily identifiable glutamatergic neurons in their tail, allowing precise quantification of neurodegeneration. In untreated UA198 worms, only 1–3 of these neurons remained intact after seven days, whereas wild-type (WT) DA1240 worms exhibited all five neurons intact (Figure 1C). Treatment with IU1 preserved neuronal integrity in most worms, maintaining numbers comparable to WT, thereby demonstrating IU1’s neuroprotective efficacy in an in vivo model of Aβ-induced neurodegeneration.

2.2. In Silico Screening and In Vitro Evaluation of IU1-Based USP14 Binding Ligands

Given the robust neuroprotective effect of IU1, we sought to develop novel IU1 analogues with potentially higher potency. AtomNet^®^ is the first deep-learning neural network for structure-based drug design and discovery, and its speed and accuracy make it the most advanced technology for small-molecule binding affinity prediction. This AI-driven virtual screening approach using the AtomNet^®^ deep-learning platform (Atomwise, San Francisco, CA, USA) was employed to identify USP14-binding ligands from extensive chemical libraries, including Mcule and Enamine [19,20,21]. Libraries were filtered using established medicinal chemistry criteria to remove pan-assay interference compounds (PAINS), aggregators, and auto-fluorescent compounds.

IU1 has a steric blockade mechanism to inhibit USP14 by binding to the thumb-palm cleft region of the USP14 catalytic domain, which prevents the substrate from binding to the enzyme. IU1 binds underneath the BL2 loop (residues 429–433) with distinct electron density in the substrate-binding cleft, 8.3 Å away from the catalytic cysteine Cys114 [23,24]. The binding site of IU1 in USP14 is snug and is unlikely to accommodate a larger, more elaborate compound. It is suggested that it would be very challenging to include additional nonpolar interactions to improve the binding affinity. Therefore, exploration of other ubiquitin binding sites might be necessary for blocking USP14 catalytic activity, but not affect its regulatory functions. This target site on the USP14 is recommended for screening since multiple high-resolution inhibitor-bound crystal structures are available, where the inhibitor compounds are positioned in this structurally conserved site. This pocket is also unique to USP14, which should result in compound specificity. In an effort to identify additional compounds, the target site will be expanded to encompass the catalytic Cys114. The top-ranked compounds were clustered using the Butina algorithm, resulting in 71 structurally diverse candidates that were synthesised for experimental validation (Table S1) [19,20,21,25,26].

These 71 compounds were tested in the MC65 cell model to assess their ability to rescue Aβ-induced cytotoxicity (viability assay), with IU1 serving as the reference compound (Figure S2). All the treatments were done on Aβ-producing cells (−Tet cells) and non-Aβ cells (+Tet cells) kept as controls. Of the 71 candidates, two compounds, designated AA10 and AA51 (Figure 1D–F), demonstrated superior cytoprotective effects, preserving cell viability above 50% relative to untreated −Tet cells. Specifically, IU1 treatment increased viability by ~53%, whereas AA10 and AA51 increased viability by ~56% and ~65%, respectively (Figure S2). Based on the cell viability screening results, AA10 and AA51 had better neuroprotective effects than the other 69 small molecules. Thus, AA10 and AA51 were chosen for further in vitro and in vivo testing.

Subsequent dose-response analyses with IU1, AA10, and AA51 (0.05–5 µM) showed dose-dependent neuroprotection, with maximal effects observed at 2.5–5 µM. Across all three compounds, 2.5 µM consistently increased viability by ~50–60%, validating this concentration for downstream mechanistic and in vivo assays (Figure 2A).

2.3. IU1 and Analogues Reduce Aβ Accumulation in AD Cell Model

To determine whether IU1 and its analogues directly modulate Aβ aggregation, we performed live-cell imaging with the luminescent conjugated oligothiophene Amytracker™ 680, which selectively binds β-sheet-rich Aβ aggregates. Amytracker is cell-permeable and a non-toxic alternative to classic amyloid dyes (such as Thioflavin T/S and Congo Red) for specialised uses in Aβ and protein aggregation and live-cell imaging research [27]. Treatments were carried out on APP-C99/Aβ-producing cells (−Tet cells), with non-APP-C99/Aβ cells (+Tet cells) kept as controls. MC65 cells were induced to express APP-C99/Aβ (−Tet) and treated with IU1, AA10, or AA51 over eight days. Aβ/APP-C99-producing cells exhibited a pronounced and sustained increase in signal intensity, ranging from ~2.0- to 2.8-fold higher than wild-type +Tet controls throughout the seven-day time course. Treatment with IU1 reduced Amytracker fluorescence by ~2-fold compared to untreated APP-C99/Aβ-producing cells. AA51 also showed a substantial reduction in Amytracker fluorescence by ~1.8 fold. AA10 showed the strongest reduction in Aβ accumulation, maintaining values close to +Tet controls across all time points (Figure 2B). Fluorescence microscopy corroborated these findings, confirming fewer intracellular Aβ aggregates in cells treated with IU1 analogues (Figure 2C).

Western blot analysis further validated these observations. IU1 reduced Aβ levels by ~45% relative to untreated −Tet cells, whereas AA10 and AA51 reduced Aβ by 60–80%, respectively. Similarly, APP-C99 precursor protein levels decreased with IU1 (~55%), AA10 (~70%), and AA51 (~50%), demonstrating that these compounds effectively target both Aβ production and accumulation (Figure 2D–G).

2.4. IU1 and Analogues Enhance Proteasome Activity

Given that USP14 inhibition is predicted to enhance proteasomal degradation [13,23], we evaluated proteasome activity in APP-C99/Aβ-producing cells, with +Tet cells used as healthy controls. Untreated −Tet cells exhibited significantly impaired UPS activity (~46% relative to +Tet controls). IU1 treatment restored activity to ~91%, whereas AA10 and AA51 elevated activity to ~125% and ~115%, respectively, nearly equivalent to +Tet controls (Figure 3A). To see if the new analogues’ neuroprotective effects were proteasome-dependent, we used MG132. MG132 is a broad-spectrum, cytotoxic proteasome inhibitor that lacks USP14 specificity rather than acting as a target-specific inhibitor, it was used as a mechanistic probe to evaluate proteasome dependency. Proteasome inhibition with MG132 abolished proteolytic activity and negated the protective effects of IU1 and analogues to around ~2–5%. MG132 co-treatment also abolished protection against cell death, reducing cell viability to levels comparable to untreated −Tet cells (Figure 3B).

IU1, AA10, and AA51 markedly reduced Aβ accumulation by ~60–80%, respectively, in APP-C99/Aβ-producing MC65 cells, whereas MG132 alone increased Aβ levels by ~30% (Figure 3C). Co-treatment of IU1, AA10, and AA51 with MG132 almost completely abolished these Aβ-lowering effects, restoring levels to those of untreated cells (Figure 3C,D). A similar pattern was observed for APP-C99, with IU1, AA10, and AA51 reducing levels by ~55%, ~90%, and ~80%, respectively, while MG132 alone or in combination with the compounds caused a ~twofold increase (Figure 3E,F). Together, these findings show that the reductions in Aβ and APP-C99 by IU1, AA10, and AA51 are proteasome-dependent.

2.5. IU1 and Analogues Restore Autophagy–Lysosomal Function

We next investigated whether IU1 analogues modulate autophagic pathways similar to the parent compound in APP-C99/Aβ-producing cells (−Tet). Western blot analysis revealed significant LC3 and p62 accumulation in untreated −Tet cells, indicative of impaired autophagy. IU1 reduced LC3I by ~35%, while AA10 and AA51 achieved reductions of ~70% and ~65%, respectively. LC3II levels decreased by ~70% with all compounds, and p62 levels were reduced by ~80–85%, restoring their levels to +Tet controls (Figure 4A,B). Immunofluorescence and LysoTracker assays showed that IU1, AA10, and AA51 decreased the elevated colocalisation of LC3/p62 and acidified vesicle (AV) accumulation, consistent with restoration of autophagy–lysosomal function. Chloroquine was used as an autophagy inhibitor control, which behaved similarly to the −Tet cells. (Figure 4C,D). Next, to determine whether Aβ co-localises with lysosomal marker, we did immunofluorescence colocalisation studies using LysoTracker coupled with Aβ labelling. In −Tet MC65 cells, we detected substantial colocalisation of Aβ with lysosomal markers. IU1 and its analogues AA10 and AA51 treatment reduced this lysosomal build-up, demonstrating that Aβ is sequestered into lysosomal vesicles and targeted for destruction via the autophagy–lysosomal pathway (Figure S3).

2.6. IU1 and Analogues Reduce Neurodegeneration and Improve Behaviour in C. elegans AD Model

In UA198 worms, untreated worms exhibited severe glutamatergic tail neuron loss, retaining only 1–3 of the five neurons after seven days. Treated UA198 worms with IU1, AA10, and AA51 preserved neuronal integrity, with 77%, 85%, and 83% of worms retaining all five neurons, respectively, compared to 35% in untreated controls. Similar to MC65 cells, LysoTracker staining indicated improved lysosomal function in Aβ-expressing UA198 worms treated with IU1 AA10 and AA51 (Figure 5A,B).

Lifespan assays showed that wild type (WT) DA1240 worms survived ~26 days, whereas UA198 worms survived ~21 days. IU1, AA10, and AA51 treatment significantly extended median lifespan to ~24, ~25, and ~25 days, respectively (Figure 5C). Behavioural assays further showed functional rescue: glycerol avoidance (“drop test”) responsiveness improved from ~45% in untreated UA198 worms to ~85% in treated groups (Figure 5D), and mechanosensation responses increased from ~75% to ~85% (Figure 5E).

2.7. MG132 Abolishes Neuroprotective Effects in C. elegans

To assess proteasome dependency in vivo, UA198 worms were co-treated with MG132 and IU1 or analogues. Neurodegeneration protection decreased from ~77% to ~61% (IU1), ~82% to ~45% (AA10), and ~83% to ~60% (AA51) (Figure 6A). In the mechanosensation assay, WT DA1240 worms showed ~90% responsiveness, while UA198 worms had only ~72% reaction. IU1-, AA10-, and AA51-treated UA198 worms improved response to ~84%, ~82%, and ~82%, respectively; however, MG132 alone reduced responsiveness by ~35%. Moreover, MG132 co-treatment lowered IU1, AA10, and AA51 mediated recovery to ~60–65%. In the drop test, DA1240 worms responded at ~87% compared with ~45% in UA198 worms, and IU1, AA10, and AA51 restored responsiveness to ~80–85%. These improvements were substantially diminished by MG132, reducing neuroprotection to ~35–40%. Overall, MG132 markedly reduced the behavioural rescue induced by all three compounds.

In summary, IU1 and its novel analogues, AA10 and AA51, consistently mitigated Aβ-mediated toxicity, enhanced proteasome activity, restored autophagy–lysosomal function, reduced Aβ and APP-C99 accumulation, and improved neuronal integrity, survival, and behaviour in both cell and C. elegans AD models. Notably, the analogues displayed superior efficacy to the parent compound IU1, highlighting the therapeutic potential of targeting the proteasomal clearance pathway for AD intervention. These effects were dependent on intact proteasome function, as shown by MG132 inhibition in both in vitro and in vivo models. Together, these findings show IU1 analogues as promising candidates for further preclinical development in AD and potentially other neurodegenerative disorders.

3. Discussion

Our research here focused on targeting the UPS as a therapeutic strategy to improve Aβ degradation. Numerous natural proteasome activators, such as oleuropein, betulinic acid, ursolic acid, AM-404, and MK-886, have been documented to confer neuroprotective benefits through the stimulation of the UPS [28,29,30,31]. Oleuropein enhances proteasomal chymotrypsin-like activity through Nrf2 activation, whereas triterpenoids such as betulinic and ursolic acid indirectly promote proteasome function via Adenosine 5′-monophosphate (AMP)-activated protein kinase/ Nuclear factor erythroid 2-related factor 2 (AMPK/Nrf2) signalling [32,33]. AM-404, an anandamide uptake inhibitor, and MK-886, a 5-lipoxygenase-activating protein inhibitor, have shown efficacy in diminishing Aβ and tau pathology via augmented proteasomal degradation and the induction of autophagy [34]. However, these compounds exhibit inadequate brain bioavailability, poor solubility, potential cytotoxicity, pleiotropic effects, and metabolic instability, hindering clinical translation. Although direct pharmacological activation of the proteasome is challenging, with only a few studies documenting this approach [35]. Recent strategies emphasise indirect modulation, including the inhibition of DUBs or modulation of regulatory signalling pathways [36].

Inhibition of DUB activity is a promising strategy for cancer therapy. Recent findings have garnered attention towards DUBs as selective regulators of proteostasis, primarily for reducing protein aggregation in neurodegenerative diseases. Among these, USP14 has emerged as a negative regulator of proteasomal activity that regulates protein ubiquitination, substrate degradation and intersects with autophagy signalling [23]. Nonetheless, USP14 is not the sole deubiquitinating enzyme associated with neurodegeneration. Malfunctions in the ubiquitin C-terminal hydrolases (UCH)-family of DUBs, particularly ubiquitin C-terminal hydrolase L1 (UCHL1) and ubiquitin C-terminal hydrolase L5 (UCHL5) through mutations, oxidative damage, or dysregulated activity, have been closely associated with neurodegenerative disorders such as AD and Parkinson disease (PD) [37,38]. Other DUBs, such as ubiquitin-specific peptidase 9, X-linked (USP9X), modulate autophagy through Raptor/mechanistic target of rapamycin (mTOR) regulation and play a role in tau homeostasis [39]. Pharmacological inhibition of USP9X enhanced autophagy and expedited the degradation of tau and phosphorylated tau. Likewise, ubiquitin-specific-processing protease 5 (USP5), ubiquitin-specific-processing protease 7 (USP7), ubiquitin-specific-processing protease 8 (USP8), ubiquitin-specific-processing protease 30 (USP30), and UCHL1 have been associated with APP processing, mitochondrial quality control, and neuroinflammatory signalling [40,41,42,43]. In particular, USP7 and ubiquitin-specific-processing protease 10 (USP10) govern the degradation of proteins associated with oxidative stress and autophagy, highlighting their extensive functions in proteostasis [44,45]. These findings collectively indicate that various DUBs converge on proteasome and autophagy pathways, and the selective targeting of regulators like USP14 offers a refined strategy for restoring proteostasis in AD.

This study identifies USP14 inhibitor IU1 and its analogues as effective modulators of proteostasis that attenuate Aβ-mediated neurotoxicity in AD models, with differential efficacy between compounds. For example, IU1-248 and IU1-47 demonstrated reduced efficacy, whereas AA10 and AA51 showed enhanced bioactivity as compared to the parent molecule IU1. A structural comparison of IU1 and its analogues (IU1-47, IU1-248, AA10, and AA51) showed that they all possess a conserved aromatic scaffold featuring a benzimidazolinone backbone essential for USP14 binding [9,23], yet they vary in their ring substitutions and side-chain orientations, which likely accounts for their differing potencies [23,24]. IU1-47 and IU1-248 were originally designed to possess superior binding affinity to USP14, given their larger substituents that occupy the hydrophobic pocket of USP14. In IU1-47, the phenyl/benzene ring features a larger substituent (chlorine instead of fluorine), which more effectively occupies the binding pocket, resulting in enhanced hydrophobic interactions. In IU1-248, the benzene ring features a CN substituent (a carbon atom triple-bonded to a nitrogen atom), which is larger and more polar than the fluorine in IU1. In summary, the larger or more polar substituents of IU1-47 and IU1-248 may lead to impaired membrane permeability or restricted cellular uptake, resulting in the observed diminished activity in the AD cell model.

Like IU1, AA10 possesses a benzylaminopyrimidine extension that augments hydrogen bonding and hydrophobic interactions, which are essential for USP14 affinity and stability, and key to better proteasomal functionality. AA51 features an additional fluorophenyl-isoindole-carbohydrazide motif that enhances cell permeability while minimising steric hindrance. Both AA10 and AA51 maintain the fundamental IU1-like aromatic framework while optimising side-chain topology and polarity to augment better proteasome engagement. This structural optimisation aligns with their enhanced efficacy in MC65 cell and C. elegans AD models, where both compounds increased Aβ/APP-C99 clearance and superior neuroprotection relative to the parent compound IU1. This study offers a proof of concept for the mechanistic basis for the systematic development of next-generation USP14 inhibitors as disease-modifying therapies for AD and associated proteinopathies.

This study focused on the modulation of proteostasis pathways by small molecules; however, further work is required to determine how these compounds influence Aβ aggregation and whether such effects translate into enhanced clearance of protein aggregates and neuroprotection. Emerging nanoscale imaging technologies can further highlight the relevance of proteostasis-based strategies. Super-resolution and nanoscale spectroscopic investigations show that Aβ aggregation goes through structurally diverse and transitory stages, with early oligomeric species causing more neurotoxicity compared with mature fibrils [46,47]. These findings suggest that therapeutic strategies targeting intracellular clearance mechanisms, including the ubiquitin–proteasome system and autophagy, may achieve neuroprotection by redirecting pathogenic aggregation processes rather than by simply lowering total Aβ load. Also, future study using ultra-resolution imaging will be highly beneficial for understanding sub-cellular localisation and Aβ deposition, which will greatly improve understanding of the pathways of Aβ clearance.

The current study has a number of limitations that should be noted. Mechanistically, our findings show that intracellular degradation pathways play a major role in Aβ clearance, without affecting Aβ production or secretion. The protective effects of IU1 and its analogues were eliminated by co-treatment with the proteasome inhibitor MG132, indicating that an intact proteasome is necessary for their function. While MG132 does not provide solid evidence of USP14 selectivity due to its wide and cytotoxic nature, our findings strongly indicate proteasome reliance of the observed neuroprotection, and off-target effects from other deubiquitinases cannot be excluded. While our findings support a proteasome-dependent mechanism for the neuroprotective effects of IU1 and its analogues, off-target contributions cannot be ruled out. IU1 has been shown to influence a variety of pathways, including the inhibition of other DUBs (such as USP7), calpain activation, modulation of ubiquitin E1 activity, induction of mitophagy, and regulation of MDM2 and mitochondrial function [48]. The extent to which these pathways contribute to the observed effects in our models is yet to be investigated. Future studies incorporating direct USP14 enzymatic assays and systematic off-target profiling will be required to delineate pathway specificity and optimise therapeutic selectivity.

In C. elegans models, treatments were not evaluated in non-amyloidogenic control worms (DA1240), preventing assessment of compound effects under normal physiological conditions. Furthermore, translation to mammalian systems is still a crucial next step, even though MC65 cells and C. elegans offer reliable and complementary discovery-phase models. Future research will extend validation to primary neuronal cultures and animal AD models, including pharmacokinetic and blood-brain barrier assessments, and examine combination tactics with autophagy modulators or immunotherapies.

4. Methods and Materials

4.1. In Silico Screening of USP14 Inhibitors

A large-scale virtual screen for small-molecule USP14 inhibitors was conducted using AtomNet^®^ (Atomwise, San Francisco, CA, USA), a deep-learning neural network for structure-based drug discovery, as part of the Artificial Intelligence Molecular Screen (AIMS) Program (A19-330, 2019) [19,20,21]. Commercially available compound libraries from Mcule and Enamine were used as starting materials. These libraries were filtered using the Eli Lilly medicinal chemistry filters [25] to eliminate chemically undesirable structures, and potential false positives, including pan-assay interference compounds (PAINS) [49], aggregators, and autofluorescent compounds were removed. The remaining compounds were screened against the USP14 target (UNIPROT: P54578). To minimise redundancy, molecules with >0.5 Tanimoto similarity (ECFP4 space) to known USP14 ligands or analogues from homologous proteins (>70% sequence identity) were excluded. The top 30,000-scoring compounds were clustered using the Butina algorithm [26], and 85 representative exemplars were purchased. Liquid chromatography-mass spectrometry data from Atomwise shows that 71 molecules met the >90% purity threshold (data not provided, commercially available compounds). These compounds were synthesised and prepared as 10 mM Dimethyl sulfoxide (DMSO, Merck, Sydney, Australia, D8418) stock solutions for further selection in the in vitro MC65 model.

4.2. Drug Screening in MC65 Cells

To identify USP14 inhibitors capable of rescuing Aβ-induced toxicity, all 71 compounds were screened in MC65 cells expressing APP-C99/Aβ. MC65 cells undergo cell death in the absence of tetracycline (−Tet, Merck, Sydney, Australia, T7660) due to accumulation of APP-C99 and Aβ, while remaining viable in its presence (+Tet). Compounds were tested in triplicate in −Tet conditions. The entire screening assay was independently repeated four times (n = 3 per replicate, four independent replicates). Compounds showing >50% protection relative to untreated −Tet controls were selected as hits for further validation.

4.3. Proteasome Activity Assay

Proteasome activity was quantified using the Proteasome Activity Assay Kit (Abcam, Adelaide, Australia, ab107921), which measures chymotrypsin-like activity via cleavage of the fluorogenic peptide Succ-LLVY-AMC. MG132 was included in parallel for all samples to inhibit proteasomal proteolysis and identify non-proteasomal background activity. Specific proteasome activity was calculated by subtracting inhibitor Relative Fluorescence Units (iRFU) from total RFU. All experiments were repeated four times (n = 3 per replicate, four independent replicates).

4.4. Immunofluorescence

MC65 cells were treated with test compounds for 72 h, washed in Phosphate buffer saline (PBS, Merck, Sydney, Australia, 11666789001), and fixed with 4% paraformaldehyde (BioScientific, Sydney, Australia, 18814-20) for 15 min at room temperature, as previously described [50]. Cells were permeabilised using 0.1% Triton X-100 (Bio-Rad, Sydney, Australia, 1610407) for 15 min, followed by blocking with 1% bovine serum albumin (BSA, Merck, A9418) for 1 h at room temperature. Primary antibodies targeting LC3 (Cell Signalling Technology, Danvers, MA, USA, 4108, 1:250) and p62 (Cell Signalling Technology, Danvers, MA, USA, 88588, 1:500) were applied overnight at 4 °C. After PBS washes, cells were incubated with Alexa Fluor-conjugated secondary antibodies anti-rabbit IgG (Cell Signalling Technology, Danvers, MA, USA, 4414, 1:5000) and anti-mouse IgG (Cell Signalling Technology, Danvers, MA, USA, 4408, 1:5000) for 1 h. Nuclei were stained with Hoechst (Cell Signalling Technology, Danvers, MA, USA, 4082S, 1 µg/mL). The laser intensity and exposure time were further calibrated based on the control autofluorescence levels, typically within 30–50% laser power, with exposure times of ~800 ms to 1 s, depending on the fluorophore. Imaging was performed using a Nikon Eclipse Ti2 fluorescence microscope (Nikon corporation, Tokyo, Japan).

4.5. Western Immunoblotting

Protein extraction, sodium dodecyl-sulfate polyacrylamide gel electrophoresis, and Western blotting were performed as previously described [51].

4.6. Cell Viability Analysis

MC65 cells were cultured in DMEM/F12 (Thermo Scientific, Melbourne, Australia, 11320082) supplemented with 10% fetal calf serum (FCS, Thermo Scientific, Melbourne, Australia, 10100147). Cells were seeded in 96-well plates at 50,000 cells per well and allowed to adhere overnight. Prior to compound treatment, media were replaced with Opti-MEM (Thermo Scientific, Melbourne, Australia, 22600134), with or without tetracycline. Cells were treated with test compounds for 72 h. Viability was measured using the MTS assay (Abcam, Adelaide, Australia, ab197010) as previously described [52].

4.7. Amytracker Aβ Aggregation Assay

Intracellular Aβ aggregates were quantified using Amytracker™ 680 (Ebba Biotech, Perth, Australia, A680-A-100). After three days of culture in −Tet conditions, cells were washed and incubated with Amytracker 680 for 30 min at 37 °C. Excess dye was removed with three washes, after which Opti-MEM containing treatment compounds was added. Fluorescence readings (Ex/Em 630/680 nm) were collected daily for eight days using a VantaStar microplate reader (BMG Labtech GmbH, Ortenberg, Germany). On day 6, fluorescence microscopy was performed using Nikon Eclipse Ti2 (Nikon corporation, Tokyo, Japan).

4.8. Lysotracker Staining (Cells and C. elegans)

In the MC65 model, cells were seeded at 20,000 per well in 96-well plates and treated for 72 h. Cells were labelled using 50 nM LysoTracker™ (Thermo Scientific, Melbourne, Australia, L12492) diluted in prewarmed Opti-MEM and incubated for 30 min at 37 °C. Cells were washed, replenished with fresh medium, and imaged under a Nikon Eclipse Ti2 microscope. In C. elegans models, synchronised adult worms (n = 20 per replicate, 3 biological replicates) were harvested on day 7 of adulthood after compound treatment. Worms were incubated in 50 µM LysoTracker diluted in prewarmed growth media for 1 h at 23 °C in darkness, allowed to recover on OP50-seeded NGM plates for 1 h, and imaged.

4.9. C. elegans Strains

Transgenic Aβ-expressing C. elegans strains DA1240 and UA198 were provided by the Caldwell Laboratory (University of Alabama, Tuscaloosa, AL, USA). All strains were maintained at 23 °C on nematode growth medium (NGM) plates seeded with E. coli OP50 following standard protocols [53].

4.10. Nematode Growth Media (NGM) Preparation

NGM was prepared by dissolving 3 g NaCl (Merck, S9888), 17 g agar (Thermo Scientific, Melbourne, Australia, 22700025), and 2.5 g peptone (MP Biomedicals, Sydney, Australia, 0210480801) in 975 mL of distilled water and autoclaving. After cooling to 55 °C, sterile supplements were added: 0.3 mL 1 M CaCl_2_ (DKSH Australia Pty Ltd., Sydney, Australia, VWRC27810.295), 0.3 mL 1 M MgSO_4_ (Sigma, Melbourne, Australia, M2643), 0.3 mL of 5 mg/mL cholesterol (dissolved in ethanol, [Sigma, Melbourne, Australia, C8667]), and 7.5 mL 1 M KPO_4_ buffer (16.282 g of K_2_HPO_4_ [Sigma, Melbourne, Australia, P3786], 0.888 g of KH_2_PO_4_ [Sigma, Melbourne, Australia, PHR1330]). Media were dispensed into plates and dried for 48–72 h. For behavioural assays, 100 µL fluorodeoxyuridine (FUDR, Sigma, Melbourne, Australia, F0503) was added per plate to prevent progeny. OP50 lawns were prepared using overnight bacterial cultures.

4.11. Maintenance of C. elegans

Worms were maintained on OP50-seeded plates using standard methods [53]. Approximately 0.1 mL OP50 culture was spotted onto small or medium plates and allowed to form a lawn overnight. Worms were then transferred at room temperature using sterilised picks.

4.12. Pharmacological Treatment of C. elegans

IU1 (100 µM, Sigma, Melbourne, Australia, I1911), AA10 (Atomwise, San Francisco, CA, USA), and AA51 (Atomwise, San Francisco, CA, USA) were dissolved in DMSO and applied to the surface of OP50-seeded NGM plates at final concentrations of 100 µM based on previous protocols [54,55,56,57]. Synchronous L4 worms were transferred to compound-containing plates for three days, then moved to FUDR-supplemented plates with the same drug treatments for behavioural, neurodegeneration, and lifespan assays.

4.13. Neurodegeneration Assay

Neurodegeneration of glutamatergic tail neurons in Aβ-expressing strains was assessed on day 7 of adulthood. L4 worms were transferred to FUDR plates and maintained until imaging. Worms were immobilised using 3 mM levamisole (Sigma, Melbourne, Australia, L0380000) and mounted on 6% agarose pads. Fluorescent glutamatergic neurons were imaged, and neurodegeneration was scored based on the presence or loss of five normal tail neurons. Each experiment included 20 worms per group, with three biological replicates.

4.14. C. elegans Lifespan Assay

Lifespan was assessed using a modified version of the Hsin and Kenyon protocol [58]. Worms were maintained on FUDR plates at 23 °C until imaging. Survival was recorded daily until all worms had died.

4.15. Mechanosensation Assay (Gentle Touch Response)

Touch response was assessed in day-7 adult worms. Using an eyelash affixed to a pipette tip, each worm was lightly stroked on the anterior and posterior regions. Each region was tested three times. A positive response was defined as backward movement (anterior touch) or forward escape (posterior touch). Sensitivity was quantified as the percentage of positive responses. Sixty worms per strain (n = 20 per replicate, four independent replicates) were analysed.

4.16. Drop Test

Chemosensory avoidance was tested using a 50% glycerol droplet. Day-7 worms were placed on NGM plates and allowed to acclimatise for 1–2 min. A glycerol droplet (1:1 glycerol to distilled water, Merck, Sydney, Australia, G5516) was placed in front of the worm’s head, and avoidance within 4 s was scored as a positive response. Sixty worms per strain were tested across three biological replicates.

4.17. Statistical Analysis

All statistical analyses were performed using GraphPad Prism 10. Comparisons between two independent groups were analysed using unpaired Student’s t-tests. For experiments involving more than two groups, one-way or two-way analysis of variance (ANOVA) was applied as appropriate. When ANOVA revealed significant effects, post hoc multiple comparison tests were conducted using Bonferroni’s and Dunnett’s tests, as specified for individual experiments. Data are presented as mean ± standard deviation (SD). A p-value < 0.001 was considered statistically significant.

5. Conclusions

Despite the approval of Aβ-targeting antibodies, AD still has significant unmet therapeutic needs. Current immunotherapies have been reported to have adverse side effects and limited clinical benefit, emphasising the need for alternative strategies that promote Aβ clearance. In this study, we demonstrate that pharmacological modulation of the ubiquitin-proteasome system using the USP14 inhibitor IU1 and its novel analogues, AA10 and AA51, enhance intracellular Aβ/APP-C99 clearance and attenuates neurotoxicity in cellular and C. elegans models of AD. Notably, AA10 and AA51 outperformed IU1 in cell-based assays, demonstrating the efficacy of rationally optimised USP14-targeting molecules. While these findings are preliminary and require validation in mammalian systems, they demonstrate that targeting proteostasis regulators may complement existing Aβ-directed therapies and inform the development of next-generation disease-modifying strategies for AD.