Therapeutic assessment of a novel mitochondrial complex I inhibitor in in vitro and in vivo models of Alzheimer's disease

Summary

Background

Despite recent approval of monoclonal antibodies that reduce amyloid (Aβ) accumulation, the development of disease-modifying strategies targeting the underlying mechanisms of Alzheimer's disease (AD) is urgently needed.

Methods

We demonstrate that mitochondrial complex I (mtCI) represents a druggable target, where its weak inhibition activates neuroprotective signalling, benefiting AD mouse models with Aβ and p-Tau pathologies. Rational design and structure‒activity relationship studies yielded mtCI inhibitors profiled in a drug discovery funnel designed to address safety, selectivity, and efficacy.

Findings

The lead compound C458 is highly protective against Aβ toxicity, has favourable pharmacokinetics, and minimal off-target effects. C458 exhibited excellent brain penetrance, activating neuroprotective pathways with a single dose. Preclinical studies in APP/PS1 mice were conducted using functional tests, metabolic assessment, in vivo³¹P-NMR spectroscopy, blood cytokine panels, ex vivo electrophysiology, and Western blotting. Chronic oral administration improved long-term potentiation, reduced oxidative stress and inflammation, and enhanced mitochondrial biogenesis, antioxidant signalling, and cellular energetics. Efficacy against Aβ and p-Tau was confirmed in human organoids.

Interpretation

These studies provide further evidence that the restoration of mitochondrial function in response to mild energetic stress represents a promising disease-modifying strategy for AD.

Funding

This research was supported by grants from NIH AG 5549-06, NS1 07265, AG 062135, UG3/UH3 NS 113776, and ADDF 291204 (all to ET); U19 AG069701 (to TK); the Alzheimer’s Association Research Fellowship grant 23AARF-1027342 (to TKON).

Research in context.

Evidence before this study

Mild inhibition of mitochondrial complex I (mtCI) with small molecules has been shown to activate neuroprotective mechanisms. Examples include resveratrol and metformin, the latter being an FDA-approved drug for Type II diabetes and a mild mtCI inhibitor. However, most mtCI inhibitors have poor selectivity and the blood-brain barrier (BBB) penetrance, suboptimum drug-like properties, and side effects. Therefore, the development of safe and efficacious mtCI inhibitors optimised for clinical applications is important.

Added value of this study

We have previously shown that treatment with tricyclic pyrone compound CP2, a selective mtCI inhibitor, was safe and efficacious in delaying the onset and development of cognitive symptoms, reducing oxidative stress, inflammation, levels of Aβ and phospho-Tau, and blocking the ongoing neurodegeneration in multiple mouse models of familial Alzheimer's Disease (AD). To generate compounds with better drug-like properties, we developed a Drug Discovery funnel with assays optimised to assess safety, selectivity, and efficacy of small molecule mtCI inhibitors in the context of AD mechanisms. Using rational design and extensive structure-activity relationship studies, we synthesised novel molecules with properties superior to CP2. A new lead compound C458 was subjected to a comprehensive pharmacology and safety assessment, and preclinical in vivo efficacy studies in APP/PS1 mice using a battery of behaviour, metabolic, and functional tests. Target engagement and activation of the neuroprotective mechanisms of action were confirmed in the mouse brain tissue. C458 treatment reduced major AD hallmarks in human brain organoids from LOAD patients. C458 has excellent brain permeability, lack of toxicity at physiologically relevant concentrations, it is highly efficacious in cellular models of AD and APP/PS1 mice and has favourable drug-like properties.

Implications of all the available evidence

These proof-of-concept data provide additional evidence that mild inhibition of mtCI represents a promising disease-modifying strategy for AD and support further development of mtCI inhibitors for clinical application. These studies emphasise the importance of mitochondria in mediating neuroprotective mechanisms and as a druggable target for age-related neurodegenerative diseases.

Introduction

Mitochondria are essential for supporting synaptic and cognitive functions. While their primary role involves generating adenosine triphosphate (ATP) through oxidative phosphorylation (OXPHOS), recent research highlights their broader involvement in regulating cellular processes such as energy status, calcium homoeostasis, metabolite production, organelle dynamics, redox balance and immune responses, among others.^1,2 Understanding the complexity of mitochondrial signalling in cellular communication opens avenues for exploring novel therapeutic strategies that improve the fundamental mechanisms of healthy ageing.^3,4 The OXPHOS machinery, located in the inner mitochondrial membrane, consists of four electron transport chain (ETC) complexes (I–IV) and ATP synthase (complex V). Mitochondrial dysfunction, characterised by reduced OXPHOS efficiency and increased reactive oxygen species (ROS) production, is linked to neurodegenerative diseases, including AD.⁵ Interestingly, recent research has shown that genetic or pharmacological inhibition of OXPHOS complexes has beneficial effects on health and lifespan across various organisms including Drosophila, Caenorhabditis elegans, mice, and humans.6, 7, 8 However, only mild inhibition of activity was beneficial, as the complete ablation of major OXPHOS subunits resulted in a severe phenotype and a shorter lifespan. Life-extending mechanisms include adaptive responses to energetic stress mediated by AMP-activated protein kinase (AMPK).⁹ This paradoxical approach is supported by the use of metformin, an FDA-approved mtCI inhibitor^10,11 broadly prescribed to the ageing population to treat type II diabetes mellitus.¹²

Previously, we identified the small molecule tricyclic pyrone compound CP2 as a mtCI inhibitor and demonstrated its safety and neuroprotective effects in mouse models with Aβ (APP/PS1, APP, PS1, 5xTgAD) and p-Tau (3xTgAD) pathologies.13, 14, 15, 16, 17 In all the studies, CP2 reduced Aβ and p-Tau accumulation, inflammation, oxidative stress, and cognitive dysfunction while improving mitochondrial function and energy homoeostasis in the brain and periphery. Treatment was safe and efficacious when administered in utero through life, at the pre- or symptomatic stages of the disease. Chronic administration of CP2 for 14 months to APP/PS1 mice was safe, resulting in blockade of ongoing neurodegeneration and cognitive protection.^18,19 However, its complex synthesis, multiple chiral centres, and limited blood-brain barrier (BBB) penetrance pose challenges for its translation to the clinic. To address these limitations, we applied rational design and extensive structure‒activity relationship (SAR) studies to develop different compounds, where C458 (cis-(N-(pyridin-4-ylmethyl))-2-(3-(m-tolyloxy)cyclohexyl)propan-1-amine) has the best drug‒like properties. The results of the proof-of-concept studies presented below demonstrate the efficacy of C458 in multiple in vitro and in vivo AD models, further supporting the feasibility of the development of safe and efficacious mtCI inhibitors to treat neurodegenerative diseases of ageing.

Methods

Ethics declaration

All experiments with mice were approved by the Mayo Clinic Institutional Animal Care and Use Committee in accordance with the National Institutes of Health's Guide for the Care and Use of Laboratory Animals and ARRIVE guidelines. IACUC protocol number A00001186.

Reagents

CP2 was synthesised by Nanosyn, Inc. (http://www.nanosyn.com), as described previously,¹⁶ and was purified using HPLC. Authentication was performed through NMR spectra to ensure the lack of batch-to-batch variation in purity. CP2 was synthesised as a free base. For in vitro experiments, CP2 was prepared as a 10 mM stock solution in DMSO. Stock aliquots of 20 μl were stored at −80 °C. The following reagents were used: DMSO (Sigma, D2650), hydrogen peroxide 30% solution (Sigma, H1009), MEM nonessential amino acids 100× (Corning, 25-025-CI), penicillin‒streptomycin (Sigma, PO781), tetracycline hydrochloride (Sigma, T7660), high-glucose DMEM (Thermo Scientific, 11995065), heat-inactivated foetal bovine serum (Sigma, F4135), RPMI-1640 (Corning, MT10041CM), DPBS 1× (Corning, 21-031-CM), sodium pyruvate (Corning, MT25000CI), and phenol red-free Opti-MEM™ I Reduced Serum Medium (Thermo Scientific, 51200038).

Cells

Human neuroblastoma MC65 cells were a gift from Dr. Bryce Sopher (University of Washington, Seattle, USA). Expression of C99 fragment in MC65 cells was confirmed by Western blot following by cell death measurements at 72 h at Tet-Off (Aβ is expressed) conditions. AMPKα1/α2 knockout mouse embryonic fibroblasts (MEFs) were a gift from Dr. Benoit Viollet (Inserm, Paris, France). The lack of AMPKα1/α2 expression was confirmed by Western blot. Human neuroblastoma SH-SY5Y cells stably transfected with Swedish mutant amyloid precursor protein (APP_SWE) were a gift from Dr. Cristina Parrado (Karolinska Institute, Sweden). Authentication of SH-SY5Y APP_SWE cells was confirmed by real time PCR. The cells were grown in high-glucose DMEM supplemented with 10% FBS, 1 mM sodium pyruvate and 1× nonessential amino acids. Primary mouse cortical neurons were cultured as described previously.¹⁸ Neurons from neonatal NTG mice (P1) were isolated and plated from individual pups with genotype established prior to experiments. All experiments were performed in neurons after 7 days in culture. All cells were incubated in 5% CO₂ at 37 °C unless otherwise noted. Cells were regularly checked and were confirmed to be negative for mycoplasma contamination.

Antibodies

The following primary antibodies were used: p-AMPK (Thr 172) (1:1000, Cell Signaling Technology, cat. # 2535, RRID: AB_331250), AMPK (1:1000, Cell Signaling Technology, cat. # 2532, RRID: AB_330331), p-acetyl-CoA carboxylase (Ser79) (1:1000, Cell Signaling Technology, cat. #11818, RRID: AB_2687505), p-GSK3β (Ser 9) (1:1000, Cell Signaling Technology, cat. # 9323, RRID: AB_2115201), GSK3β (1:1000, Cell Signaling Technology, cat. # 9832, RRID: AB_10839406), Sirt 1 (D1D7) (1:1000, Cell Signaling Technology, cat. # 9475S), Sirt3 (1:1000, Cell Signaling Technology, cat. # 5490, RRID: AB_10828246), Superoxide Dismutase 1 (1:1000, Abcam, cat. # ab16831, RRID: AB_302535), p-Akt (Ser473) (1:1000, Cell Signaling Technology, cat. # 4051, RRID: AB_331158), and p-Akt (Thr308) (1:1000, Cell Signaling Technology, cat. # 4056, RRID: AB_331163), p-FOXO1A (Cell Signaling Technology, cat. # 84192, RRID: AB_2800035), p-FOXO3A (Cell Signaling Technology, cat. # 9466, RRID: AB_2106674), p-ULK1 (Ser757) (1:1000, Cell Signaling Technology, cat. # 14202, RRID: AB_2665508), p-ULK1 (Ser555) (1:1000, Cell Signaling Technology, cat. # 5869, RRID: AB_10707365), p-ULK1 (Ser317) (1:1000, Cell Signaling Technology, cat. # 12753, RRID: AB_2687883), PGC1α (4C1.3) ST1202 (1:1000, Calbiochem, cat. # KP9803), PGC-1α (3G6) (1:1000, Cell Signaling Technology, cat. #2178), OXPHOS cocktail (Abcam, ab110413, RRID: AB_2629281), NF-kB Pathway Sampler Kit (Cell Signaling Technology, 9936T, RRID: AB_561197), p-NF-κB p65 (Ser536) (1:1000, Cell Signaling Technology, cat. # 3033, RRID: AB_331284), IκBα (1:1000, Cell Signaling Technology, cat. # 4812, RRID: AB_10694416), HO-1 (1:1000, Cell Signaling Technology, cat. # 70081, RRID: AB_2799772), NRF2 (D1Z9C) (1:1000, Cell Signaling Technology, cat. # 12721S), Oxidative Stress Defence Cocktail (1:250, Abcam, cat. # ab179843), OxPhos Blue Native WB Antibody Cocktail (1:1000, Thermo Fisher Scientific, cat. #45-7999), mouse monoclonal NQO1 (A180) (Santa Cruz, 1:200, sc-32793, RRID: AB_628036), ND1 (USBiological, cat. # 224423), ND5 (USBiological, cat. # 224422), NDUFA9 (Thermo Fisher Scientific cat. # 459100, RRID: AB_10376187), BDNF (1:1000, Sigma–Aldrich, cat. # SAB2108004), Anti-β-Actin (1:5000, Sigma–Aldrich, cat. # A5316, RRID: AB_476743), tubulin (Rockland, cat. # 200-301-880, RRID: AB_2611065), and Complex I immunocapture kit (ab109711). The following secondary antibodies were used: donkey anti-rabbit IgG conjugated with Horseradish Peroxidase (1:10000 dilution, GE Healthcare UK Limited, UK) and sheep anti-mouse IgG conjugated with horseradish peroxidase (1:10000 dilution; GE Healthcare UK Limited, UK).

Rational design and synthesis of a mtCI inhibitor C458 (cis-(N-(pyridin-4-ylmethyl))-2-(3-(m-tolyloxy)cyclohexyl)propan-1-amine)

A complete step-by-step synthesis of C458 is presented in a protocol described by Trushin et al., Synthesis of cis-(N-(pyridin-4-ylmethyl)-2-(3-(m-tolyloxy)cyclohexyl)propan-1-amine) dx.doi.org/10.17504/protocols.io.3byl4wd98vo5/v1, 2025. C458 was synthesised by Nanosyn, Inc. (http://www.nanosyn.com), and purified via high-performance liquid chromatography (HPLC). The authentication was performed using NMR spectroscopy and HPLC-MS to ensure the lack of batch-to-batch variation in purity (>99%). For in vitro experiments, C458 was prepared as a 10 mM stock solution in DMSO. Stock aliquots of 20 μl were stored at −20 °C.

Synthesis of C458 constructs for mtCI pull-down

C458 constructs were synthesised via the C458 route, step 9, using amino alkylpyridines with 2- and 8-atom spacers attached to the NH₂ group.

Treatment of primary neuronal cultures with CP2 and C458

Primary mouse cortical neurons were cultured as previously described.¹³ Neurons from neonatal animals (P1) were isolated and plated from individual pups; genotyping was performed prior to the day of the experiment. All experiments were performed in neurons cultured for 7 days. On day 7, cells were treated with either C458 or CP2 at concentrations ranging from 1 to 20 μM for 24 h. As a control, cells were treated with vehicle (0.01% DMSO). Cell viability was measured after 24 h of treatment via the MTT assay.

Mitochondrial isolation and measurement of electron transport chain (ETC) complex activity

Intact brain mitochondria were isolated from mouse brain tissue via differential centrifugation with digitonin treatment.¹³ The brain tissue was immersed in ice-cold isolation medium (225 mM mannitol, 75 mM sucrose, 20 mM HEPES-Tris, and 1 mM EGTA, pH 7.4) supplemented with 1 mg/ml BSA. The tissue was homogenised with 40 strokes by the “B” (tight) pestle of a Dounce homogeniser in 10 ml of isolation medium, diluted twofold and transferred into centrifuge tubes. The homogenate was centrifuged at 5900×g for 4 min in a refrigerated (4 °C) Beckman centrifuge. The supernatant was centrifuged at 12,000×g for 10 min, the pellets were resuspended in the same buffer, and 0.02% digitonin was added. The suspension was homogenised briefly with five strokes in a loosely fitted Potter homogeniser and centrifuged again at 12,000×g for 10 min, then gently resuspended in isolation buffer without BSA and washed once by centrifuging at 12,000×g for 10 min. The final mitochondrial pellet was resuspended in 0.1 ml of washing buffer and stored on ice. The activity of mtCI was measured spectrophotometrically via a plate reader (SpectraMax M5, Molecular Devices, USA) in 0.2 ml of standard respiration buffer composed of 125 mM sucrose, 25 mM Tris-HCl (pH = 7.5), 0.01 mM EGTA, and 20 mM KCl at 25 °C. The NADH-dependent activity of complex I was assayed as oxidation of 0.15 mM NADH at 340 nm (ε340 nm = 6.22 mM⁻¹ cm⁻¹) in assay buffer supplemented with 10 μM cytochrome c, 40 μg/ml alamethicin, and 1 mM MgCl₂ (NADH media). NADH:Q reductase activity was measured in NADH media containing 2 mg/ml BSA, 60 μM decylubiquinone, 1 mM cyanide and 5–15 μg protein per well.

The activity of the ETC complexes was measured via complex I (Cayman Chemicals, cat. # 700930), complex II/III (Cayman Chemicals, cat. # 700950), complex IV (Cayman Chemicals, cat. # 700990), and complex V (Cayman Chemicals, cat. # 701000) colourimetric assay kits.

Pull-down of mtCI with immobilised C458 analogues

Both C458 analogues with 2-atom (458-2) or 8-atom (458-8) linkers were immobilised on Pierce NHS-activated agarose (Thermo Scientific™ Pierce™, cat. # 26197) at two concentrations, 1 mM and 0.2 mM, in 0.01 M borate buffer (pH 9) for 1 h, followed by quenching with 1 M ethanolamine overnight. NHS-activated agarose treated with 1 M ethanolamine overnight served as control beads for nonspecific binding. mtCI was pulled down with 458-2- or 458-8-immobilised agarose beads or immunocaptured with mtCI antibodies (Ab109711, Abcam) following the manufacturer's protocol. Briefly, freshly isolated mouse liver mitochondrial pellets were resuspended to 5.5 mg/ml in PBS containing protease inhibitors. The mitochondria were lysed with 1% n-dodecyl β-D-maltoside (Sigma, D4641) for 30 min on ice and then centrifuged for 10 min at 25,000×g at 4 °C. For mtCI pull-down or immunocapture, the mitochondrial lysate (1 mg in 200 μl) was incubated overnight at 4 °C with 80 μl of solid beads (either 458-2 or 458-8) or 25 μl of immunocapturing beads. Ethanolamine-immobilised beads (80 μl) were used as a negative control. After incubation, the beads were washed three times with 1 ml of lysis buffer. After washing, the proteins were eluted from the beads by heating at 95 °C for 5 min in 2× Laemmli sample buffer (Bio-Rad, cat. # 1610737). Proteins were resolved by SDS‒PAGE and immunoblotted for the ND2, ND5, and NDUFA9 subunits of mtCI.

Real-time respirometry

The kinetic injection experiment was performed via the Agilent Seahorse XFe96 Extracellular Flux Analyser. Prior to the assay, the media was replaced with Agilent Seahorse XF DMEM, pH 7.4 (Agilent, 103575-100), supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose at 37 °C for 1 h and placed in a BioTek Cytation 5 Cell Imaging Multimode Reader (Agilent). During the assay run, the compounds were injected via one of the ports, and the oxygen consumption rate (OCR) was measured every 18 min after compound injection for 1 h.

Mitochondrial stress test

This assay was performed to determine mitochondrial bioenergetics. Before each assay, the media was exchanged for Agilent Seahorse XF DMEM pH 7.4 (Agilent, 103575-100) supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose at 37 °C for 1 h without CO₂. Additionally, brightfield images were obtained via a BioTek Cytation 5 Cell Imaging Multimode Reader (Agilent). The oxygen consumption rate (OCR) was analysed under basal conditions and after treatment with different drugs, including the ATP synthase inhibitor oligomycin A (the optimal dose chosen after the dose‒response optimisation assay: 0.5–3 μM oligomycin A), an ETC uncoupler FCCP (the optimal dose chosen after the dose‒response optimisation assay: 0.5–3 μM FCCP), and an ETC inhibitor mixture (0.5 μM rotenone and 0.5 μM antimycin A). The response to the minimal dose of oligomycin A (2 μM) and FCCP (2 μM), generating the maximal effect, accounts for nonphosphorylating mitochondrial respiration and maximal FCCP-uncoupled respiration, respectively. The response to the rotenone and antimycin A mixture accounts for nonmitochondrial oxygen consumption. Hoechst 3342 dye (final concentration of 5 μM) was injected at the end of the assay, and the mixture was incubated at 37 °C for 15–30 min before fluorescence imaging via XF Imaging and Cell Counting Software for normalisation to the cell count. The spare respiratory capacity (SRC) was calculated as the difference between the maximal and basal OCRs.

Cell viability assays

MC65 cells were treated with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) at 0.25 mg/ml, followed by 4 h of incubation at 37 °C. The solubilisation of MTT formazan was performed by the addition of 200 μl of 40 mM HCl in isopropanol (2:1) in 100 μl of media for 10 min on a shaker, and the absorbance was measured at 570 nm. In some cases, cell viability was measured via a CellTiter-Glo® 2.0 assay according to the instructions provided by the manufacturer (Promega, WI).

In vitro safety pharmacology studies

In vitro safety pharmacology assays, including C458 binding to human receptors and ion channels, enzyme inhibition, and uptake measurements, were conducted by the Contract Research Organization (CRO) Eurofins (France). C458 was assessed at 10 μM (Supplementary Table S5). Compound binding was calculated as the % inhibition of the binding of a radioactively labelled ligand specific for each target. The compound enzyme inhibition effect was calculated as a percentage of the control enzyme activity. In each experiment and if applicable, the respective reference compound was evaluated concurrently with the test compound, and the data were compared with historical values determined at Eurofins. The experiment was performed in accordance with the Eurofins validation SOP. The results showing an inhibition (or stimulation for assays run under basal conditions) greater than 50% are considered to represent significant effects of the test compounds. The results showing an inhibition (or stimulation) between 25% and 50% are indicative of weak-to-moderate effects. The results showing an inhibition (or stimulation) lower than 25% are not considered significant and are mostly attributable to variability of the signal around the control level. Low-to-moderate negative values have no real meaning and are attributable to variability of the signal around the control level. High negative values (≥50%) that are sometimes obtained with high concentrations of test compounds are generally attributable to nonspecific effects of the test compounds in the assays. CRO Nanosyn, Inc. (Santa Barbara, CA) was contracted to conduct kinome wide panel (KWP) screening against 250 kinases via the established SOP. C458 was assessed at 1 μM and 10 μM concentrations. The data are provided in Supplementary Table S1.

In vivo C458 pharmacokinetic studies

The C458 pharmacokinetic profile was determined in C57BL/6J female mice ordered from the Jackson Laboratory. The mice were acclimatized for one week to the new environment prior to the initiation of the experiments. To evaluate C458 concentrations in plasma, mice were injected with a single intravenous dose of 3 mg/kg in 20% PEG 400 and 5% dextrose water via the lateral tail vein via a 0.5 cc tuberculin syringe. Independent cohorts of mice were administered C458 by oral gavage (25 mg/kg in 20% PEG 400 and 5% dextrose water) with a 1 cc tuberculin syringe with a stainless steel 22-gauge straight feeding needle. At 0, 0.08, 0.25, 0.5, 1, 2, 4, 8, and 24 h after treatment, the mice were anaesthetised, and 200 μl of blood was collected from the retro-orbital sinus through a K₂EDTA-coated capillary into a K₂EDTA-coated microtainer tube. The plasma was separated via centrifugation at 4 °C (10,000 rpm for 3 min), transferred to a microcentrifuge tube, immediately frozen on dry ice, and stored at −80 °C until analysis. The pharmacokinetic parameters of C458 were estimated via standard noncompartmental analysis. Pharmacokinetic parameters were analysed using a RS/1 computer system. Area under the curve (AUC) was measured by conventional trapezoidal summation and extrapolation.

C458 quantification via LC‒MS/MS

The method used to quantify C458 in plasma and brain tissue is described in the protocol: Trushin et al., C458 quantification using LC‒MS/MS, dx.doi.org/10.17504/protocols.io.eq2ly62xmgx9/v1, 2025.

Studies in mice

Double transgenic APP/PS1 mice were used in the study.¹⁴ The genotypes were determined via PCR as described previously.¹⁴ All the animals were maintained on a 12–12 h light‒12 h dark cycle, with a regular feeding and cage-cleaning schedule. The cages were housed in the same room to minimise potential cofounders. The mice were randomly assigned to study groups based on their age and genotype. Two trials were conducted within a few months to ensure that mice of the same age and sex were consistently treated. The first trial included three male and three female mice in the vehicle-treated group and four male and three female mice in the C458-treated group. The second trial included five male and seven female mice in the vehicle-treated group and eight female and two male mice in the C458-treated group. Data analysis was conducted on samples combined from both trials. Males and females were analysed together with the total of 18 vehicle-treated and 16 C458-treated mice. Each data point represents a single mouse. Where possible, analyses were conducted by the investigators blinded to the treatment.

Chronic C458 treatment in presymptomatic APP/PS1 mice

APP/PS1 male and female mice were given C458 (25 mg/kg/day in 0.05% DMSO dissolved in drinking water ad libitum) or vehicle-containing water (0.05% DMSO) starting at 2.5 months of age as we described previously.¹⁴ Water consumption and weight were monitored weekly to ensure mice did not reach the humane endpoint of a 20% decrease in body weight. Independent groups of mice were continuously treated for 7 months until the age of 10 months. Prior to the beginning of treatment, the mice were subjected to a battery of behaviour tests. At the end of the treatment, the mice were subjected to a behaviour battery, metabolic cages (CLAMS), and in vivo ³¹P-NMR imaging. After the mice were sacrificed, tissue and blood were collected from each mouse. The brain tissues were used for Western blot analysis, and the plasma was used for cytokine/chemokine profiling. Three mice from each group were subjected to electrophysiology recording as described below. One female mouse treated with C458 lost 20% of its body weight during the last two weeks of the study. Since this mouse did not demonstrate any signs of motor or behavioural distress, it was not removed from the study. One male and two female mice treated with C458 were found dead between 4 and 6 months of age. These mice did not exhibit weight loss or any signs of toxicity. The cause of death was not determined. All data points obtained from each mouse were included into the analyses.

Behaviour battery

Behavioural tests were carried out in the light phase of the circadian cycle, with at least 24 h between each assessment, as we described previously.¹⁴ More than one paradigm ran within 1 week. However, no more than two separate tests were run on the same day. Behavioural and metabolic tests were performed in the order described in the experimental timeline.

Open field test

Spontaneous locomotor activity was measured in brightly lit (500 lux) Plexiglas chambers (41 cm × 41 cm) that automatically recorded the activity via photobeam breaks (Med Associates, Lafayette, IN). The chambers were in sound-attenuating cubicles and were equipped with two sets of 16 pulse-modulated infrared photobeams to automatically record X–Y ambulatory movements at a 100 ms resolution. Data were collected over a 10-min trial with 30 s intervals. N = 18 for vehicle-treated APP/PS1 and n = 13 for C458-treated APP/PS1 mice.

Hanging bar

Balance and general motor function were assessed using a hanging bar. The mice were lowered onto a parallel rod (D < 0.25 cm) placed 30 cm above a padded surface. The mice were allowed to grab the rod with their forelimbs, after which they were released and scored for success (pass or failure) in holding onto the bar for 30 s. The test consisted of a 3-day trial period with three 30-s measurements taken each day. The final latency to fall in sec was presented as an average of nine tests per animal. N = 18 for vehicle-treated APP/PS1 and n = 14 for C45-treated APP/PS1 mice.

A novel object recognition (NOR) test was used to estimate memory deficits. All trials were conducted in an isolated room with dim light in Plexiglas boxes (40 cm × 30 cm). Each mouse was placed in a box for 5 min for the acclimatization period. Thereafter, a mouse was removed, and two similar objects were placed in the box. Objects with various geometric shapes and colours were used in the study. A mouse was returned to the box, and the number of interrogations of each object was automatically recorded by a camera placed above the box for duration of 10 min. A mouse was removed from the box for 5 min, and one familiar object was replaced with a novel object. Each mouse was returned to the box, and the number of interrogations of novel and familiar objects was recorded for 10 min. Experiments were analysed via NoldusЕthoVision software. The number of interrogations of the novel object was divided by the number of investigations of the familiar object to generate a discrimination index. Intact recognition memory produces a discrimination index of 1 for the training session and a discrimination index greater than 1 for the test session, which is consistent with greater interrogation of the novel object.

Morris water maze

Spatial learning and memory were investigated by measuring the time it took each mouse to locate a platform in opaque water identified with a visual cue above the platform. The path taken to the platform was recorded with a camera attached above the pool. Each mouse was trained to find the platform during four training sessions per day for three consecutive days. For each training session, each mouse was placed in the water facing away from the platform and allowed to swim for up to 60 s to find the platform. Each training session started by placing a mouse in a different quadrant of the tank. If the mouse found the platform before 60 s had passed, the mouse was left on the platform for 30 s before being returned to its cage. If the animal had not found the platform within 60 s, the mouse was manually placed on the platform and left there for 30 s before being returned to its cage. A day of rest was followed by the day of formal testing. N = 6 mice per each group.

Inflammatory markers

After 16 h of fasting, blood from APP/PS1 mice treated with vehicle or C458 was collected via orbital bleeding and centrifuged for 5 min at 5000×g. Collected plasma was sent for 32-plex cytokine array analysis (Discovery Assay, Eve Technologies Corp. https://www.evetechnologies.com/discovery-assay/). Multiplexing analysis was performed via the Luminex 100 system (Luminex). More than 80% of the targets were within the detectable range (signals from all samples were higher than the lowest standard). Undetectable targets were excluded. Blood samples were run in duplicate. N = 17 for vehicle-treated APP/PS1 and n = 13 for C458-treated APP/PS1 mice. Samples below level of quantification were excluded.

Lipid peroxidation assay

The levels of malondialdehyde (MDA), a product of lipid degradation that occurs because of oxidative stress, were measured via an MDA assay kit (#MAK085, Sigma Aldrich) in hippocampal brain tissue isolated from APP/PS1 mice treated with vehicle or C458 according to the manufacturer's instructions. N = 6 mice per each group.

Electrophysiology

APP/PS1 mice aged ∼7 months and treated with C458 (n = 3 per group) or vehicle (n = 3 per group) for ∼5 months were used for electrophysiology analysis. The mice were deeply anaesthetised with isoflurane and decapitated. The brain was quickly removed and transferred to cold slicing solution containing artificial cerebrospinal fluid (ACSF), where NaCl was substituted with sucrose to avoid excitotoxicity. Transverse slices (300–350 μm thick) were made via a vibratome (VT-100S, Leica). Slices were incubated in ACSF containing 128 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 2 mM CaCl₂, and 1 mM MgSO₄ aerated with 95% O₂/5% CO₂. Slices were maintained at 32 °C for 13 min and then maintained at room temperature throughout the entire experiment. For electrophysiology recording, each slice (2–3 slices per mouse) was transferred to a recording chamber, and ACSF was continuously perfused at a flow rate of 2–3 ml/min. A single recording electrode and a single bipolar stimulation electrode were placed on top of the slice. A boron-doped glass capillary (PG10150, World Precision Instruments) was pulled with a horizontal puller (P-1000, Sutter Instrument) and filled with ACSF for extracellular recording. Under a microscope (FN-1, Nikon), the recording electrode was placed in the CA1 area of the hippocampus. The bipolar stimulation electrode (FHC) was placed at Schaffer collaterals. The distance between the two electrodes was greater than ∼200 μm. To define the half response of the stimulation, various intensities of electrical stimulation were applied (10–500 μA). However, the pulse width was fixed at 60 μs. Once the stimulation parameter was determined to generate half the maximum evoked fEPSP, this stimulation intensity was used for paired pulse and LTP experiments. For the LTP experiment, test stimulation was applied every 30 s for 30 min to achieve a stable baseline. Once a stable baseline was achieved, tetanic stimulation (100 Hz for 1 s) was applied three times with 3-s intervals. The initial slope of the fEPSP was used to compare the synaptic strength.

Comprehensive Laboratory Animal Monitoring System (CLAMS)

CLAMS (Columbus Instruments, Columbus, OH) is a set-up allowing automated, non-invasive and simultaneous monitoring of horizontal and vertical activity, feeding and drinking, oxygen consumption and CO₂ production of an individual mouse. APP/PS1 mice treated with vehicle or C458 (9–10 months old) were individually placed in CLAMS cages. Indirect calorimetry was monitored over 2 days, where the mice were allowed food for 24 h ad libitum (fed state), and for the following 24 h, the food was removed (fasting state). The mice were maintained at 20–22 °C under a 12:12 h light–dark cycle. All the mice were acclimatized to CLAMS cages for 3–6 h before recording. Sample air was passed through an oxygen sensor for determination of the oxygen content. Oxygen consumption was determined by measuring the oxygen concentration in the air entering the chamber compared with that in the air leaving the chamber. The sensor was calibrated against a standard gas mixture containing defined quantities of oxygen, carbon dioxide and nitrogen. Food and water consumption were measured directly. The hourly file displayed measurements for the following parameters: VO₂ (volume of oxygen consumed, ml/kg/h), VCO₂ (volume of carbon dioxide produced, ml/kg/h), RER (respiratory exchange ratio), heat (Kcal/h), total energy expenditure (TEE, kcal/h/kg of lean mass), activity energy expenditure (AEE, kcal/h/kg of lean mass), resting total consumed food (REE kcal/h/kg of lean mass), food intake (g/kg of body weight/12 h), metabolic rate (kcal/h/kg), total activity (all horizontal beam breaks in counts), ambulatory activity (minimum 3 different, consecutive horizontal beam breaks in counts), and rearing activity (all vertical beam breaks in counts). EE, RER, and fatty acid (FA) oxidation were calculated via the following equations: RER = VCO₂/VO₂; EE (kcal/h) = [3.815 + 1.232 × RER] × VO₂ × 1000; and FA oxidation (kcal/h) = EE × (1 − RER/0.3). Daily FA oxidation was calculated from the average 12 h of hourly FA oxidation. Daily carbohydrate plus protein oxidation was calculated from the average 12-h hourly EE minus daily FA oxidation. Metabolic flexibility was evaluated from the difference in RER between the daily fed state and fasted state recorded at night according to the following equation: Δ = 100% ∗ (RER fed-RER fasted)/RER fed. N = 9 for vehicle-treated APP/PS1 and n = 7 for C458-treated APP/PS1 mice.

Dual-energy X-ray absorptiometry (DEXA)

The LUNAR PIXImus mouse densitometer (GE Lunar, Madison, WI), a dual-energy supply X-ray machine, was used for measuring skeletal and soft tissue masses for the assessment of skeletal and body composition in C458- or vehicle-treated mice. Live mice were scanned under 1.5–2% isoflurane anaesthesia. The mice were individually placed on plastic trays, which were then placed onto the exposure platform of the PIXImus machine to measure body composition. The following parameters were generated: lean mass (in grams), fat mass (in grams), and the percentage of fat mass. These parameters were used to normalise the data generated in CLAMS, including O₂, VCO₂, metabolic rate and energy expenditure. N = 9 for vehicle-treated APP/PS1 and n = 7 for C458-treated APP/PS1 mice.

In vivo³¹P-NMR spectroscopy

NMR spectra were obtained from 10-month-old APP/PS1 mice treated with C458 (n = 6 per group) or vehicle for ∼7 months (n = 7 per group) via an Avance III 300/700 MHz (7/16.4 T) wide-bore NMR spectrometer equipped with microimaging accessories (Bruker, Billerica, MA, USA) with a 25-mm inner diameter dual nucleus (³¹P/¹H) birdcage coil. For anatomical positioning, a pilot image set of the coronal, sagittal, and axial imaging planes was used. For ³¹P spectroscopy studies, a single pulse acquisition with a pulse width of P1 200 ms (∼30°), spectral width of SW 160 ppm, FID size of TD 16k, FID duration of AQT 0.41 s, waiting time of D1 1 s and number of scans of NS 512 was used. The acquisition time was 12 min. Spectra were processed via TopSpin v3.5 software (BrukerBiospin MRI, Billerica, MA). The integral areas of the spectral peaks corresponding to inorganic phosphate (Pi), phosphocreatine (PCr), and the γ, α, and β phosphates of adenosine triphosphate (αATP, βATP, and γATP) were measured. Since the Pi peaks were not detectable in some mice, the Pi values were uniformly omitted from the analyses. The presence of phosphomonoester (PME) or phosphodiester (PDE) peaks was also recorded. However, the signal‒to‒noise ratios of these peaks are not always adequate for accurate quantification. The levels of PCr and Pi were normalised by the total ATP levels present in that spectrum or by the amount of βATP. The results were consistent between both normalisation methods; the data are presented as the ratio of each parameter to total ATP.

Differentiation of iPSCs into cortical organoids

Cortical organoids were generated via the use of the STEMdiff™ Cerebral Organoid Kit (Stemcell Technologies) according to the manufacturer's instructions, with slight modifications. On day 0, the iPSCs were dissociated into single-cell suspensions with TrypLE Express (Thermo Fisher Scientific, Waltham, MA, USA) and cultured on U-bottom ultralow-attachment 96-well plates (15,000 cells/well) in embryoid body (EB) formation media (medium A) supplemented with 10 μM Y27632. After the medium was changed to EB formation media on days 2 and 4, the iPSC-derived EBs were transferred onto 96-well low-attachment plates on day 5 and cultured in induction medium (medium B). On day 7, the EBs were transferred onto Matrigel-coated 6-well plates and cultured in expansion medium (medium C + D) for cortical organoid formation for 3 days. On day 10, the culture medium was changed to maturation medium (medium E). After 4 weeks, medium E was replaced with neuronal maturation medium, which was composed of DMEM/F12 + neurobasal medium (1:1) supplemented with N2, B27, BDNF (20 ng/ml), GDNF (20 ng/ml), ascorbic acid (200 μM), and Dibutyryl Cyclic AMP (Db-cAMP) (100 nM) (Sigma–Aldrich). The organoids were cultured on an orbital shaker, and the medium was changed twice per week until day 60.

Organoid processing

Organoids were collected 48 h after treatment with compounds and lysed in RIPA lysis buffer supplemented with protease and phosphatase inhibitor cocktails (Roche). The lysed samples were sonicated at 40% amplitude. The samples were subsequently centrifuged at 21,000×g for 45 min at 4 °C. The total protein concentration in the soluble fraction was quantified via the Pierce BCA Protein Assay Kit.

ELISA quantification

The levels of Aβ40, Aβ42, total tau, and phospho-Tau were measured via the following ELISA kits: the Human β-Amyloid (1–40) ELISA Kit (Thermo Fisher, KHB3481), the Human β-Amyloid (1–42) ELISA Kit (Thermo Fisher, KHB3544), the Human Tau ELISA Kit (Thermo Fisher, KHB0041), and the Tau (Phospho) [pT181] Human ELISA Kit (Thermo Fisher, KHO0631), according to the manufacturers' instructions. The samples and detection antibodies were added to the ELISA plates and incubated for the specified duration of time. Following incubation, the wells were washed four times with wash buffer and incubated with an IgG-HRP conjugate for the indicated time. After additional washes, the plates were treated with stabilised chromogen and incubated for 30 min at room temperature in the dark. The reaction was stopped with a stop solution, and the absorbance was measured at 450 nm via a microplate reader (BioTek). The results were normalised to the total protein concentration of the cell lysate.

Western blot analysis

Protein levels in the cortico-hippocampal region of the brains of vehicle- or C458-treated APP/PS1 mice were determined via Western blot analysis. The tissue was homogenised and lysed via RIPA buffer (25 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) containing phosphatase PhosSTOP (Roche, cat. # 04906837001) and protease inhibitors (cOmplete, Roche, cat. # 11697498001). Total protein lysates (25 μg) were separated on 4–20% Mini-PROTEAN TGX™ Precast Protein Gels (Bio-Rad, 4561093). For the C458 time course study, coronal brain slices that encompassed the cortico-hippocampal region were homogenised, and 30 μg of protein lysates were separated on 4–15% Criterion gels (Bio-Rad, cat. # 5678083) and transferred to Immun-Blot polyvinylidene difluoride membranes (PVDF cat. # 1620177). Total cell lysates were prepared using RIPA buffer. Bands were imaged and quantified via a KwikQuant imager and KwikQuant image analyser 5.9 (Kindle Biosciences, USA).

γ-Secretase-mediated peptide cleavage assay

Human neuroblastoma SH-SY5Y cells grow to confluence (∼2 × 10⁸ cells per sample). Cells were harvested with a scraper and washed and pelleted with ice cold D-PBS centrifuged at 400×g for 4 min. Cell pellet or mouse liver were resuspended in buffer A (20 mM HEPES, 50 mM KCl, 2 mM EGTA, and complete protease inhibitor mixture, pH 7.5) with the ratio of 1:8 and lysed using 15 strokes of a Dounce homogeniser. After homogenisation, lysates were centrifuged at 800×g for 10 min at 4 °C to remove nuclei and large debris. Microsomal membrane fractions were pelleted from the postnuclear supernatants by centrifugation at 21,000×g for 2.5 h at 4 °C. Membrane pellets were resuspended (0.5 mg/ml) in 20 mM Hepes, pH 7.0, 150 mM KCl, 2 mM EGTA, 1% (w/v) CHAPSO (Calbiochem), and protease inhibitor mixture and solubilised at 4 °C for 1 h with end-over-end rotation. The solubilised membranes were centrifuged at 21,000×g for 1 h, and the supernatants were collected. For measuring γ-secretase activity, solubilised membranes were resuspended in buffer B (50 mM Tris-HCl, 2 mM EGTA, 150 mM NaCl, 0.25% CHAPSO) with 8 μM γ-secretase substrate (565764, Calbiochem, pH 6.8), compounds were added, and the solution was incubated at 37 °C overnight. The solubilised membranes were treated with vehicle, γ-secretase inhibitor L685,458 (10 μM), C458 (250 or 500 nM), or CP2 (250 or 500 nM). Fluorescence was measured at λex = 355 nm; λem = 440 nm using Varioskan™ LUX Multimode Microplate Reader (Thermo Fisher Scientific). Analysis and presentation of results were conducted using Microsoft Excel and Graph Prism software.

Statistics

All the statistical analyses were performed via GraphPad Prism 10.5.0. The statistical analysis included two-tailed unpaired and paired Student's t tests (where appropriate) and one-way ANOVA. The power calculation was based on a one-way ANOVA comparing up to four groups, assuming a power of 0.7, a large effect size (Cohen's f ≈ 0.4, equivalent to d ≈ 0.75), and a significance level of 0.05. Based on these parameters, a minimum of five animals per group was estimated. The study was not powered to assess sex-specific differences. When P values were significant at a level of P < 0.05, Fisher's LSD post hoc analysis was applied to determine the differences among groups. The data are presented as the means ± SDs for each group of mice.

Role of funders

This research is solely the responsibility of the authors and does not necessarily represent the official view of the NIH. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Results

Design and synthesis of C458

To improve the drug-like properties of the neuroprotective mtCI inhibitor CP2 (Fig. 1a), we used a rational design to develop series of small molecules. These compounds were screened using a drug-discovery funnel developed and optimised in the laboratory to prioritise the best candidates (Fig. 2a). The primary assays evaluated the cytotoxicity, potency, selectivity as mtCI inhibitors, and target engagement across a broad concentration range (Fig. 2b–e). Among the 24 tested compounds, C458 (cis-(N-(pyridin-4-ylmethyl))-2-(3-(m-tolyloxy)cyclohexyl)propan-1-amine) (Fig. 1a) exhibited the most favourable drug-like properties (Table 1).

Fig. 1.

Rational design and chemical synthesis of C458. a The rational design of C458 was based on the CP2 structure, with structural modifications applied to rings A, B, and D. b The chemical synthesis of C458 involved the conversion of 5-cis (IV) to 5-trans (VI), resulting in a cis isomer C458 (X) with an improved yield of 51%.

Fig. 2.

C458 is a nontoxic mild mtCI inhibitor effective against Aβ toxicity. a Drug Discovery Funnel for C458 identification. b C458 and CP2 do not cause cell death in WT mouse cortical neurons at concentrations less than 10 μM (24-h MTT assay). P values were determined by one-way ANOVA. c C458 protects MC65 Tet-Off cells (Aβ expressed, red line) with an EC₅₀ of 4.6 ± 0.6 nM. No toxicity was observed in MC65 Tet-On cells (Aβ not expressed, black line). EC₅₀ values were calculated by fitting experimental data to calculated data for nonlinear regression via GraphPad Prism 10 software. The data are expressed as the means ± SDs, n = 3 technical replicates. d C458 and rotenone increase NADH levels in MC65 Tet-On cells, reflecting mtCI inhibition. NADH was measured over 24 h. The data are expressed as the means ± SDs. e Compared with complete inhibition by 1 μM rotenone (NADH oxidation assay), CP2, and C458 mildly inhibited mtCI activity in isolated mouse brain mitochondria. The data are expressed as the means ± SDs, n = 4 technical replicates. f C458 reduces the oxygen consumption rate (OCR) in MC65 Tet-On cells less potently than rotenone does. The quantified data are shown in Supplementary Fig. S1h. g Pretreatment with rotenone blocks the protective effect of C458 in MC65 cells expressing Aβ. Open black diamond: Tet-On cells (no Aβ expression) treated with vehicle, serving as a positive control (100% viability). Open blue triangle: Tet-Off condition (Aβ expressed), resulting in significant cell death after 3 days. Blue line: Tet-Off cells treated with C458 (2 nM–5 μM) show full protection from Aβ toxicity, maintaining ∼100% viability. Solid red triangle: Tet-On cells treated with 32 nM rotenone show no toxicity, confirming this concentration is not inherently cytotoxic. Open red triangle: Tet-Off cells treated with 32 nM rotenone alone exhibit complete cell death, indicating sensitivity to rotenone under Aβ-expressing conditions. Red line: pre-treatment with 32 nM rotenone blocks the protective effect of C458 in Tet-Off cells, supporting a competitive interaction at mtCI. Partial rescue at higher C458 concentrations reflects incomplete occupancy of the mtCI binding site by rotenone, which inhibits ∼50% of mtCI activity at this dose. The data are expressed as the means  ± SDs, n = 4 technical replicates. d–eP values were calculated via unpaired Student's t tests. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗∗P < 0.0001. All experiments were reproduced in two or more biological replicates. #, EC₅₀ for C458.

Table 1.

Drug-like properties of CP2 and C458.

	CP2	C458
Calculated log P	3.0	4.4
Calculated log D7.4	2.0	2.0
Total polar surface area	109	34.2
H-bond acceptors	8	3
H-bond donors	2	1
Molecular weight	393.45	338.50

Compared with CP2, C458 has fewer hydrogen donors and acceptors, abridged stereochemistry with an open structure, fewer aliphatic carbons, and structural modifications to the A, B, and D rings (Fig. 1a). Additionally, its synthesis was simplified (Fig. 1b). By isolating the unwanted cis-isomer (IV) and converting it into the trans-isomer (VI) via a two-step sequence, the yield of the third synthetic step improved from 30% to 51%. All subsequent experiments utilised the cis-C458 isomer (X).

The cytotoxicity and potency of C458 (Fig. 2a–c) were evaluated in the human neuroblastoma MC65 cell line, which expresses the C99 fragment of the amyloid precursor protein (APP) under the control of a tetracycline (Tet)-sensitive promoter.²⁰ When tetracycline is withdrawn (Tet-Off), MC65 cells express C99, which is converted to Aβ by γ-secretase. This leads to cell death within three days due to Aβ accumulation, causing oxytosis, ferroptosis, and mitochondrial dysfunction, key mechanisms involved in AD pathogenesis.²¹ In the presence of tetracycline (Tet-On), MC65 cells behave normally and can be used to assess compound cytotoxicity (Fig. 2a, Assay 1). These cells provide an excellent phenotypic screening assay for drug discovery (Fig. 2a, Assay 2). The efficacy of CP2 against Aβ toxicity was initially demonstrated via this assay.²² Like CP2, C458 did not induce cytotoxicity up to 10 μM in primary mouse cortical neurons (Fig. 2b) or in MC65 Tet-On cells (Fig. 2c, black line). Furthermore, treatment of MC65 Tet-Off cells with C458 prevented Aβ-induced cell death, with an EC₅₀ of 4.6 ± 0.6 nM (Fig. 2c, red line), compared with 150 nM for CP2.²² These results show that C458 provides superior protection against Aβ-induced mechanisms compared with CP2, with no cytotoxicity across a wide concentration range, demonstrating excellent safety.

Target validation of C458

The target engagement (Fig. 2a, Assay 3) and selectivity of C458 for mtCI inhibition (Fig. 2a, Assay 4) were evaluated by measuring the enzymatic activities of complexes I–V in isolated mouse brain mitochondria (Fig. 2e, Supplementary Fig. S1a–c). At the efficacious concentrations (0.002 μM–5 μM) established in Assay 2, C458 weakly inhibited the activity of mtCI (Fig. 2e and f) without affecting the activities of the other complexes (Assay 4, Supplementary Fig. S1a–c). At 1 μM, C458 reduces mtCI activity by approximately 15%, in contrast to rotenone, which completely inhibits mtCI at the same concentration (Fig. 2e). Since mtCI inhibition by C458 at efficacious concentrations was very mild, we measured NADH accumulation in cells over time because of reduced NADH oxidation (Fig. 2d).²³ NADH accumulation in MC65 Tet-On cells after C458 treatment occurred within the same concentration range required to rescue MC65 cells from Aβ toxicity, confirming target engagement (Fig. 2c and d). This inhibition resulted in a moderate ∼20% decrease in ATP production (Supplementary Fig. S1d). The weak inhibition of mtCI by C458 was further confirmed by a slower rate of NADH accumulation in MC65 cells than rotenone treatment (Fig. 2d). As a second measure, differences in the extent of mtCI inhibition were evaluated by measuring the oxygen consumption rate (OCR) in MC65 Tet-On cells following kinetic injections of C458 or rotenone via a Seahorse Extracellular Flux Analyser (Fig. 2f). While rotenone immediately blocked respiration, C458 gradually and dose-dependently reduced the basal OCR (Fig. 2f, Supplementary Fig. S1f). At C458 concentrations of 1 μM or lower, changes in the OCR were minimal, and there was no increase in the extracellular acidification rate (ECAR), indicating the absence of increased glycolysis (Supplementary Fig. S1e). Only at concentrations above 10 μM did C458 significantly decrease the OCR and increase the ECAR (Supplementary Fig. S1e and f). Thus, at nanomolar concentrations, C458 weakly inhibits mtCI, protecting MC65 cells from Aβ toxicity without causing cytotoxicity or inducing glycolysis, closely mimicking the mechanisms of CP2.¹³

Our recent data (manuscript in preparation), generated via Cryo-electron microscopy (Cryo-EM) and purified ovine mtCI, confirmed that CP2 binds to the deep quinone site of mtCI (Qd site), one of three binding sites for rotenone.^24,25 Since C458 was designed on the basis of the CP2 structure (Fig. 1a), we hypothesise that it also binds to the quinone cavity. To test this hypothesis, we developed a competitive inhibition assay using rotenone to block the Qd site on mtCI. MC65 cells were pretreated with 32 nM rotenone for 30 min before C458 treatment. This rotenone concentration did not induce cell death in MC65 cells but effectively reduced the OCR by 50% within 30 min of rotenone addition (Supplementary Fig. S1g and h). Importantly, rotenone significantly reduced the protective activity of C458 against Aβ toxicity at a single nanomolar dose (EC₅₀, Fig. 2g) and markedly diminished its effectiveness at higher concentrations (Fig. 2g). Treatment with any concentration of rotenone did not protect MC65 cells against Aβ toxicity (data not shown). These findings suggest that mild inhibition of mtCI by C458 because of its binding to the Qd-site of mtCI is crucial for its protective effect.

To further confirm C458 binding to mtCI, we developed a pull-down assay using C458 derivatives modified with spacers of varying lengths, each terminating with an NH₂ group for immobilisation onto functionalized agarose beads (Fig. 3a). Spacer-modified constructs (C458-8 with an 8-atom spacer and C458-2 with a 2-atom spacer) showed the same efficacy against Aβ toxicity in MC65 cells as unmodified C458, indicating that the spacers did not affect the activity of C458 (Supplementary Fig. S2a). In the pull-down assay, C458-8 immobilised at higher concentrations (1 mM) on agarose beads efficiently captured mtCI from mouse brain mitochondrial lysates. This efficiency was greater than that of C458-2 or C458-8 immobilised at a lower concentration (0.2 mM) (Supplementary Fig. S2b). The ability of C458-8-immobilised beads to capture mtCI was comparable to that of mtCI immunocapture via specific antibodies (Fig. 3b, IC). The specificity of the interaction was further demonstrated by the lack of nonspecific binding to control beads immobilised with ethanolamine (Fig. 3b, control).

Fig. 3.

C458 penetrates the BBB, directly interacts with mtCI, and activates neuroprotective pathways in the mouse brain. a C458 analogues were designed with two- and eight-atom spacers based on SAR studies. b Pull-down assay using C458-8 immobilised on agarose bead isolates mtCI from mitochondrial lysates. c, d Pharmacokinetics of C458 in plasma after intravenous (IV) and oral (PO) administration in C57BL/6 female mice (n = 3 per time point). e C458 levels in the brain after oral administration (25 mg/kg via gavage) in C57BL/6 female mice, demonstrating BBB penetration. f Activation of neuroprotective pathways in mouse brain tissue after a single oral dose of C458 (25 mg/kg). Western blot analysis was performed on brain tissues collected at various time points post administration. Each lane represents an individual mouse.

To assess off-target effects, C458 activity was evaluated via a 250-kinase panel provided by Nanosyn, Inc. Similar to CP2,¹⁴ C458 did not inhibit any of the 250 kinases at concentrations of 1 μM or 10 μM (Supplementary Table S1). Taken together, these findings confirm that C458 is a specific mtCI inhibitor that protects against Aβ toxicity in a cellular model of AD.

C458 penetrates the BBB and activates neuroprotective mechanisms in the brain

The pharmacokinetic (PK) profile of C458 was evaluated in female wild-type mice (WT, C57BL/6) by administering 3 mg/kg intravenously (IV) or 25 mg/kg orally (PO) (Table 2). Blood and brain tissues were collected from each mouse over a 24-h period (Fig. 3c–e), and C458 levels in frozen plasma and brain samples were measured via LC‒MS/MS. C458 demonstrated high BBB penetrance, with a brain/plasma ratio of 3.18, and oral bioavailability of 47.4% (Table 2). These indicate that C458 has favourable distribution between plasma and brain essential for the development of CNS drugs. A brain/plasma ratio above 3 implies that therapeutically relevant brain concentrations can be achieved without requiring excessively high plasma exposure, potentially reducing peripheral side effects. Following PO administration, the half-life (T_1/2) of C458 was 2.6 h in plasma and 6.8 h in the brain, with a maximum concentration (C_max) of 1819.6 ng/ml reached within 1 h (Fig. 3c and d; Table 2). BBB penetrance was further confirmed by an efflux ratio (ER = 1.1) in Madin–Darby canine kidney (MDCKII) cells transfected with the human MDR1 gene encoding p-glycoprotein (Supplementary Table S2).

Table 2.

The pharmacokinetics and bioavailability of C458 in female C57BL6 mice.

Summary	Dose	T(1/2)a	Tmaxb	Cmaxc	Clastd	AUClaste	AUCinff	Keg	Clh	F (%)i	Brain/plasma ratio
	(mg/kg)	hr	hr	ng/ml	ng/ml	ng/ml∗h	ng/ml∗h		ml/hr/kg
IV plasma	3	4.21	0.08	775.33	1.02	695	701	0.16	4.28
PO plasma	25	2.61	0.25	1819.57	1.05	2768	2772	0.27	9.02	47.39
PO brain	25	6.77	1.00	1705	57.02	8266	8823	0.10	2.83		3.18

Summary of PK study of C458 delivered to C57BL/6 female mice (n = 3 per time point) via IV route (3 mg/kg) or orally (PO route) (25 mg/kg via gavage).

^aHalf-life.

^bTime to maximum concentration.

^cMaximum concentration.

^dLast measurable concentration.

^eArea under the curve from the last measurable concentration.

^fArea under the curve extrapolated to infinity.

^gElimination rate constant.

^hClearance.

ⁱBioavailability.

To assess target engagement and activation of neuroprotective mechanisms, Western blot (WB) analysis was performed using brain tissues from the PK study. Within 1 h of PO administration, C458 activated AMPKα, a key regulator of cellular energy homoeostasis, resulting in the downstream inactivation of acetyl-CoA carboxylase 1 (ACC1) (Fig. 3f).²⁶ Consistent with AMPK activation, the phosphorylation of Unc-51, similar to Autophagy Activating Kinase 1 (ULK-1), at S317 and S555 indicated autophagy activation (Fig. 3f). Additionally, C458, like CP2,13, 14, 15 promoted Akt phosphorylation (glucose metabolism) and inactivated Gsk3β and FOXO3A (tau phosphorylation and neuroprotection), engaging multiple neuroprotective mechanisms²⁷ (Fig. 3f). These effects persisted for 24 h post administration. These findings indicate that C458 efficiently crosses the BBB, promptly engages mtCI, and activates a neuroprotective integrated signalling cascade.¹⁸

To evaluate in vivo toxicity, independent groups of 2-month-old male and female WT mice were treated with vehicle or 25 mg/kg or 50 mg/kg C458 via the drinking water ad libitum. One month later, the adult and newborn mice were sacrificed, and the liver, spleen, heart, lungs, kidneys, sex organs, and brain were subjected to histopathological examination. No developmental or tissue pathology was observed in C458-treated mice (Supplementary Table S3). Thus, similar to CP2,¹³ C458 showed no developmental toxicity or adverse effects after 30 days of treatment at 50 mg/kg/day.

C458 treatment alleviates the AD-like phenotype in APP/PS1 mice

The preclinical in vivo efficacy of C458 was evaluated in two independent cohorts of APP/PS1 mice aged 2.5 to 10.5 months (Fig. 4a). Trial 1 assessed the safety of C458 administration and its impact on behaviour, cognitive function, and inflammation, whereas Trial 2 focused on reproducibility and effects on metabolism, brain and peripheral energy homoeostasis, synaptic function, and mechanisms of action. Data from tests applied in both trials were combined for analyses.

Fig. 4.

C458 treatment improves motor and cognitive functions and LTP in APP/PS1 mice. a Timeline of two trials in which male and female APP/PS1 mice were treated with either vehicle or C458 for 7.5 months, with tests conducted in each trial. b Body weights of APP/PS1 mice treated with vehicle or C458 over the study duration (data from Trials 1 and 2 combined). c C458 treatment reduces (a trend) hyperactivity in APP/PS1 mice in the open field test. d C458 treatment enhances motor strength and coordination in the hanging bar test. e, f C458 treatment improved performance in the novel object recognition test and in the Morris water maze. Data for Trials 1 and 2 are shown in c–e, n = 18 vehicle-treated and n = 13 C458-treated mice per group. Data from Trial 2 are shown in f, n = 6 mice per group. g C458 treatment improved basal synaptic strength. The relationships between the initial slopes of fEPSPs and presynaptic fibre volley amplitudes were significantly greater in the C458-treated APP/PS1 group than in the vehicle-treated APP/PS1 group. h, i C458 treatment enhances LTP in APP/PS1 mice; n = 2–3 slices from 3 mice per group. Representative traces are shown as the mean ± SDs at each time point. g–i Data from Trial 2. The data were analysed via unpaired Student's t test, except for the NOR test, which was analysed via paired Student's t tests. ∗P < 0.05, ∗∗P < 0.01.

As the study was not designed to evaluate sex-specific differences, data from male and female mice were analysed together. Non-transgenic littermates were not included in this study. APP/PS1 mice received 25 mg/kg/day C458 or vehicle (0.05% DMSO) via ad libitum drinking water. The brain and plasma C458 concentrations used in the present study ranged from 37 to 1000 nM, which is consistent with mtCI engagement based on the PK data (Supplementary Table S4). C458-treated APP/PS1 mice exhibited no toxicity or side effects and gained weight throughout the study (Fig. 4b).

Compared with vehicle treatment, functional tests revealed significant improvements in motor and cognitive performance in C458-treated APP/PS1 mice. Compared with their vehicle-treated counterparts, the C458-treated APP/PS1 mice demonstrated reduced hyperactivity in the open field test, enhanced motor coordination and muscle strength in the hanging bar test (Fig. 4c and d), improved attention and nonspatial declarative memory in the novel object recognition test, and better spatial memory and learning in the Morris water maze (Fig. 4e and f). Given that synaptic loss strongly correlates with cognitive dysfunction in AD,^28,29 synaptic function was examined to investigate the basis of improved cognitive performance. Extracellular recordings of field excitatory postsynaptic potentials (fEPSPs) were obtained from the CA1 region of hippocampal slices acutely prepared from APP/PS1 mice treated with either C458 or vehicle for 5 months. Compared with vehicle treatment, C458 treatment increased synaptic strength in APP/PS1 mice and significantly improved long-term potentiation (LTP) (Fig. 4h and i). These findings indicate that chronic C458 treatment is safe and significantly improves cognitive and behavioural AD-like phenotypes in APP/PS1 mice.

C458 treatment enhances metabolic flexibility and energy homoeostasis in the brain and periphery

CP2-treated APP/PS1 mice presented improved energy homoeostasis in the brain and periphery, as we reported previously.¹⁴ To investigate the effect of C458 on metabolism, indirect calorimetry (CLAMS) was employed. Compared with their vehicle-treated counterparts, APP/PS1 mice treated with C458 presented increased carbohydrate oxidation and metabolic flexibility, as indicated by the respiratory exchange ratio (RER), a critical marker of the ability to adapt to changes in metabolic or energy demands consistent with improved glucose utilisation (Fig. 5a–e).³⁰ C458 treatment also significantly improved glucose tolerance in APP/PS1 mice (Fig. 5f), suggesting enhanced glucose metabolism. Since C458 inhibits mtCI, we examined whether chronic C458 administration affects ATP levels in the brain.

Fig. 5.

C458 treatment enhances metabolic function and glucose tolerance and preserves ATP in the brains of APP/PS1 mice. a, b Changes in the respiratory exchange ratio (RER) over 44 h during ad libitum feeding for all treatment groups. c, d C458 treatment increased glucose oxidation during ad libitum feeding, as shown by the CLAMS data. The grey bars represent fat consumption; the orange bars represent carbohydrate and protein oxidation. e Metabolic flexibility tended to improve in C458-treated APP/PS1 mice, as evidenced by their ability to switch from carbohydrate metabolism to fat metabolism between the feeding and fasting states. a–en = 9 vehicle-treated and 7 C458-treated mice per group. f C458 improves glucose tolerance, as assessed by the intraperitoneal glucose tolerance test (IPGTT). g Representative in vivo³¹P-NMR spectra comparing vehicle-treated (blue) and C458-treated (red) APP/PS1 mice. h Phosphocreatine/ATP ratio calculated from ³¹P-NMR spectra, demonstrating ATP preservation in C458-treated APP/PS1 mice, n = 7 vehicle-treated and 6 C458-treated mice per group. The data in a, b, f, g, and h are presented as the means ± SDs; those in c–e are presented as the means ± SEMs. Statistical analysis was performed via Unpaired Student's t test, except for c, d, which was analysed using two-way ANOVA. ∗∗P < 0.01, ∗∗∗∗P < 0.0001. The data shown are from Trial 2.

We utilised ³¹P nuclear magnetic resonance (³¹P-NMR) spectroscopy, a non-invasive, in vivo method for real-time measurements of key energy metabolites, including phosphocreatine (PCr), inorganic phosphate (Pi), and the α, β, and γ phosphate groups of ATP (Fig. 5g and h).³¹ This translational method is also used to assess energy levels in humans, providing crucial insights into brain energy dynamics. After 7 months of C458 treatment, APP/PS1 mice tended toward an increased PCr/ATP ratio. In cells, PCr serves as a critical energy reservoir, buffering ATP levels during periods of high energy demand.³¹ The observed trend toward increase in the PCr/ATP ratio suggests improved maintenance of energy reserves (Fig. 5g), and indicates that mild inhibition of mtCI by C458 doesn't impair mitochondrial function and energy efficiency within brain cells. Collectively, these findings suggest that C458 treatment promotes metabolic adaptation, thereby preserving or even enhancing energy homoeostasis in the brain while also improving glucose metabolism in the periphery, indicating that C458 could help regulate systemic energy homoeostasis.

C458 treatment mitigates oxidative stress and inflammation

AD is associated with increased oxidative stress and ROS generation, altering redox homoeostasis.^5,32 Nicotinamide adenine dinucleotide phosphate (NADPH) is a crucial molecule for maintaining the cellular redox balance and functions as an electron donor to reduce glutathione (GSH), a major cellular antioxidant.³³ In AD mouse neurons, NADPH depletion occurs upstream of GSH exhaustion, contributing to oxidative stress and promoting neuronal death.³⁴ Similarly, the expression of Aβ in MC65 cells induces ROS production, depleting GSH and triggering oxytosis/ferroptosis-induced cell death.^21,35 To assess whether C458 provides protection against oxidative stress following Aβ accumulation in MC65 cells, we challenged MC65 cells with different doses of hydrogen peroxide (H₂O₂) after pretreatment with C458 or vehicle for 24 h (Fig. 6a). The expression of Aβ significantly sensitised cells to H₂O₂-induced oxidative stress, reducing their survival (Fig. 6a). This was associated with a significant decrease in NADPH and GSH levels compared with those in MC65 Tet-On cells (Fig. 6b and c). C458 treatment restored resistance to H₂O₂ to a level comparable to that observed in MC65 Tet-On cells (Fig. 6a). Consistently, C458 significantly increased NADPH and GSH levels in MC65 Tet-Off cells in a dose-dependent manner (Fig. 6d and e). These findings highlight that restoring NADPH and GSH levels via C458 treatment is crucial for mitigating Aβ-induced oxidative stress and ferroptosis in MC65 cells.

Fig. 6.

C458 treatment increases resistance to oxidative stress and improves redox balance in MC65 Tet-Off cells. a C458 treatment increases resistance to oxidative stress caused by H₂O₂ treatment in MC65 cells expressing Aβ (Tet-Off condition). LD50 values were calculated through nonlinear regression via GraphPad Prism software. The data are expressed as the means ± SDs. b, c Expression of Aβ (Tet-Off condition) in MC65 cells significantly depletes NADPH (b) and GSH (c) pools. d, e C458 restores NADPH (d) and GSH (e) levels in MC65 Tet-Off cells. The data are presented as the means ± SD. f Levels of malondialdehyde (MDA) were reduced in the brain tissue of C458-treated APP/PS1 mice; n = 6 mice per group. g–j C458 reduces the levels of proinflammatory cytokines and chemokines in the plasma of APP/PS1 mice. g, IL-1β; h, INF-γ; i, TNFα; j, MIP-2. n = 17 vehicle-treated and 13 C458-treated mice per group. Data from Trials 1 and 2 combined. Statistical analysis was performed via unpaired Student's t tests, except for d, e, which used one-way ANOVA. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001.

We next measured lipid peroxidation, a well-established marker of oxidative stress, in the brain.³⁶ After chronic C458 treatment in APP/PS1 mice, the levels of malondialdehyde (MDA), a measure of lipid peroxidation, tended to decrease (Fig. 6f), indicating that C458 treatment effectively mitigates oxidative stress in both in vitro and in vivo models of AD. Aβ accumulation in MC65 cells depletes GSH, increases ROS and lipid peroxidation, and leads to NF-kB activation and the synthesis of proinflammatory cytokines.^21,37 To determine whether the reduction in oxidative stress after C458 treatment alleviates inflammation,³⁸ we measured the levels of 32 cytokines and chemokines in the plasma of APP/PS1 mice (Fig. 6g–j). C458 significantly decreased the plasma levels of the proinflammatory cytokines IL-1β, TNFα, and IFN-γ, as well as the chemokine MIP-2, indicating that the treatment alleviates chronic inflammation in APP/PS1 mice (Fig. 6g–j).

Neuroprotective mechanisms of C458 require AMPK activation

Our previous studies suggested that the efficacy of mtCI inhibitors in AD models requires AMPK activation.^13,14 To confirm that AMPK is essential for C458-dependent neuroprotective mechanisms, we used AMPKα1/α2 knockout mouse embryonic fibroblasts (MEFs) (a generous gift from Dr. B. Violet).³⁹ In these MEFs, the absence of AMPK results in constitutive activation of the NF-κB inflammatory pathway, as evidenced by increased degradation of IκBα, a key inhibitor of NF-κB, and subsequent nuclear translocation of the transcription factor p65 NF-κB.⁴⁰

In wild-type (WT) MEFs, C458 activates AMPKα, leading to inactivation of its downstream target ACC and an increase in the levels of antioxidants, including haem oxygenase-1 (HO-1, Thioredoxin and Nrf-2), as well as elevated levels of IκBα (Fig. 7a). These findings indicate that C458 suppresses inflammation through the AMPK-mediated inhibition of NF-κB and enhanced antioxidant capacity. Furthermore, in WT MEFs, C458 stimulates mitochondrial biogenesis, as evidenced by increased levels of peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) and OXPHOS complexes I–V (Fig. 7a, Supplementary Figs. S3 and S7). In contrast, vehicle-treated AMPKα1/α2-deficient MEFs presented decreased baseline levels of PGC1α, OXPHOS complexes I, II, and V, HO-1, NQO-1, and IκBα (Fig. 7a, Supplementary Fig. S3a and b, Supplementary Fig. S7a and b). In AMPKα1/α2-deficient MEFs, C458 failed to promote mitochondrial biogenesis, antioxidant defences, or anti-inflammatory responses, demonstrating that AMPK is essential for the activation of neuroprotective adaptive stress response pathways.

Fig. 7.

C458 exerts its neuroprotective effects through AMPK activation, which initiates multiple protective pathways in both in vivo and in vitro models. a C458 treatment activated AMPKα and PGC1α, increased the levels of OXPHOS complexes and antioxidants, and mitigated NF-κB signalling in WT MEFs but not in AMPKα1/α2-deficient MEFs. b In primary cortical mouse neurons, C458 (2.5 μM) treatment activated AMPKα in a time-dependent manner and induced mitochondrial biogenesis, as evidenced by PGC1α activation and increased levels of OXPHOS complexes I–V. c In SH-SY5Y APP_SWE cells, C458 treatment decreased the basal OCR while increasing the maximal OCR compared with terminal OCR inhibition via rotenone (256 nM). d C458 treatment improved SRC in SH-SY5Y APP_SWE cells. The SRC was calculated for the OCR data shown in (c). e In the brains of APP/PS1 mice (Trial 2; n = 6 mice per group), C458 treatment activates AMPKα, induces mitochondrial biogenesis via PGC1α activation, reduces GSK3β and FOXO1/3A activity, and enhances autophagy and the levels of BDNF and Sirtuins 1 and 3. The data are presented as the means ± SDs. Statistical analysis was performed via one-way ANOVA to compare the vehicle- and C458-treated groups. ∗P < 0.05; ∗∗P < 0.01.

Consistent with the observed increase in mitochondrial biogenesis and the levels of OXPHOS complexes in MEFs, C458 treatment increased PGC1α and the levels of OXPHOS complexes in primary mouse neurons over a time course of 72 h (Fig. 7b, Supplementary Figs. S4 and S8). Additionally, C458 enhanced bioenergetics in human neuroblastoma SH-SY5Y cells stably transfected with Swedish mutant amyloid precursor protein (APP_SWE), another cellular AD model characterised by increased Aβ production⁴¹ (Fig. 7c and d). Specifically, compared with vehicle-treated cells, C458-treated cells presented increased spare respiratory capacity (SRC), an indicator of mitochondrial fitness and the ability to produce energy under stress conditions. Notably, treatment with rotenone, a terminal mtCI inhibitor, did not improve SRC (Fig. 7c and d). This enhanced SRC aligns with the improved energy homoeostasis observed following C458 treatment in APP/PS1 mice.

To confirm the in vivo neuroprotective mechanisms of C458, WB analysis was performed on brain tissue from APP/PS1 mice treated with vehicle or C458 for 7.5 months (Fig. 7e). C458 treatment reached efficacious concentrations in mouse brains (Supplementary Table S4) and activated key neuroprotective pathways, including the AMPKα, ULK-1, and PGC1α pathways. It also inactivated critical metabolic regulators, such as ACC, Gsk3β, and the transcription factors FOXO1A/FOXO3A (Fig. 7e, Supplementary Figs. S5 and S9). We also observed increases in the levels of sirtuin (SIRT) 1 and 3 and brain-derived neurotrophic factor (BDNF), indicating enhanced support for neuronal function. These findings demonstrate that, like CP2, C458 activates multiple neuroprotective mechanisms in vitro and in vivo, reducing inflammation and augmenting cellular bioenergetics, redox balance, and mitochondrial function.

To explore C458 as a potential drug candidate, a comprehensive safety assessment was conducted via the Cerep Safety Screen 44 panel, which includes various human receptors, neurotransmitters, and ion channels. At a concentration of 10 μM, which is 1000× greater than the EC₅₀ concentration used in efficacy Assay 2 (Fig. 2a), C458 inhibited several human receptors, including hERG (Supplementary Table S5), a critical potassium channel subunit essential for maintaining normal cardiac electrical activity. Taking into consideration the structural differences between human hERG and murine mERG, this potential cardiac liability may not be accurately detected in preclinical in vivo studies using mice. This highlights the importance of introducing hERG safety assessments earlier in the drug discovery pipeline to guide the development of selective, potent, and safe mtCI inhibitors.

C458 reduces the levels of Aβ and p-Tau in iPSC-derived cerebral organoids from patients with sporadic AD

To assess the translational potential of mtCI inhibitors, we utilised iPSC-derived cerebral organoids from patients with AD carrying two APOE4 alleles, which represent sporadic AD.⁴² These well-characterised 3D organoids exhibit an APOE4/4-associated phenotype, including elevated levels of Aβ, p-Tau T181, apoptosis, and synaptic loss.⁴² Following 48 h of treatment with C458, the organoids were lysed in RIPA buffer, and the levels of Aβ40, Aβ42, p-Tau T181, and total tau were quantified via ELISA⁴² (Supplementary Fig. S10). C458 treatment significantly reduced the intracellular Aβ42/Aβ40 ratio and the p-Tau/Tau ratio (Fig. 8a–f) without affecting γ-secretase activity (Supplementary Fig. S11a and b). These results indicate that mild mtCI inhibition mitigates Aβ and Tau pathology in human brain cells, which is consistent with data generated in 3xTgAD and APP/PS1 mice treated with C458 or CP2.^14,15 Furthermore, the reduction in the intracellular Aβ42/Aβ40 ratio is particularly notable given that APOE4 variants are linked to impaired amyloid clearance and increased Aβ42 accumulation. Collectively, these findings demonstrate that mtCI inhibition ameliorates key AD hallmarks in iPSC-derived cerebral organoids from patients with sporadic AD.

Fig. 8.

Treatment with C458 reduces the levels of Aβ and p-Tau T181 in iPSC-derived organoids from patients with sporadic AD. a–c The levels of Aβ40, Aβ42, and Aβ42/Aβ40 in RIPA fractions from human iPSC-derived organoids treated with vehicle or C458 were measured by ELISA. d–f The levels of p-Tau T181, total tau, and pTau/Tau in RIPA fractions from human iPSC-derived organoids treated with vehicle or C458 were measured by ELISA. The data were normalised to the total protein concentration of each sample (n = 3 organoids per treatment, measured in duplicate). The data are expressed as the means ± SDs. Statistical analysis was performed via Student's t tests or one-way ANOVA. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

Discussion

The development of disease-modifying treatments for AD is hindered by an incomplete understanding of the diverse mechanisms that underlie heterogeneous and varied responses to interventions.^43,44 While lecanemab,⁴⁵ a disease-modifying anti-amyloid antibody, has been approved by the FDA, its side effects raise concerns with its application. Alternative approaches targeting the underlying mechanisms of AD pathobiology must be considered. Targeting mtCI with small molecules capable of restoring multiple mechanisms essential for AD represents a promising strategy.¹⁸

We validated the feasibility of using safe and efficacious mtCI inhibitors for AD treatment. Building on the previously characterised mtCI inhibitor CP2, we developed a small molecule, C458, with improved drug-like properties; PK; and absorption, distribution, metabolism, and excretion (ADME) profiles. C458 demonstrated strong neuroprotective effects in both in vitro and in vivo models of AD. C458 reduced Aβ and p-Tau levels in human brain organoids generated from the iPSCs of patients with sporadic AD, emphasising its translational potential. C458 exhibited potency against Aβ toxicity at nanomolar concentrations, with no detectable cytotoxicity in neuronal cells up to 10 μM, and safe application in APP/PS1 mice over 7 months, indicating a good safety profile. C458 effectively crosses the BBB and provides neuroprotection in APP/PS1 mice within a therapeutic brain concentration range of 100–1000 nM. The concentration of C458 in the brains of APP/PS mice and the brain/plasma ratio were much greater than those reported for CP2,¹⁴ indicating significantly better BBB penetration, a lack of excessive efflux, and enhanced stability. Compared with the recently developed mtCI inhibitors for cancer therapy, C458 has an outstanding safety profile. Unlike phenformin and IACS-01759, which can induce lactic acidosis due to strong inhibition of mtCI and enhanced glycolysis,^46,47 C458 mildly inhibits mtCI at efficacious doses. Its mechanism of action leads to a gradual accumulation of NADH, increasing by up to 25% at 1 μM over time (Fig. 2d). This mild and sustained mtCI inhibition results in a modest 10–15% reduction in ATP levels, which is sufficient to activate AMPK via an elevated AMP/ATP ratio.⁴⁸ The activation of AMPKα through mtCI inhibition by C458 mirrors the effects of exercise⁴⁹ and calorie restriction,⁵⁰ where the degree and timing of inhibition influence the mode of AMPK activation.⁵¹ A single oral dose of C458 induced robust AMPKα activation comparable to that induced by high-intensity exercise (Fig. 3f). Chronic C458 treatment, which was administered ad libitum for more than 7.5 months, produced mild and sustained AMPK activation (Fig. 7e), akin to long-duration, moderate-intensity exercise. Both high- and moderate-intensity exercise preferentially activate AMPKα2,⁵² suggesting that the energetic stress induced by C458 inhibition of mtCI likely also activates this isoform. Moreover, C458 may selectively activate mitochondria-associated AMPK, the mitoAMPK α2/β2/γ1 isoform, which responds first to energetic stress induced by mtCI inhibition or exercise.⁵³ This selective activation of mitoAMPK, particularly the regulation of mitophagy and biogenesis, is critical for mitochondrial quality control.⁵³ Given that the AMPKα2 isoform plays a vital role in regulating long-term synaptic plasticity and memory formation,⁵⁴ C458 may preferentially activate the mitoAMPK α2/β2/γ1 isoform in neurons, providing an additional layer of selective targeting of cells with high energy demand.

In contrast to complete mtCI inhibition by rotenone, weak mtCI inhibition by C458 does not increase ROS production or trigger inflammation. In contrast, C458 treatment restored the NADPH/GSH pools, effectively halting Aβ-mediated toxicity and preventing oxytosis/ferroptosis–mediated death in MC65 cells. It also reduces oxidative stress and chronic inflammation in APP/PS1 mice. The increase in the NADPH pool is consistent with the inhibition of ACC1 and ACC2 by AMPKα, which decreases NADPH consumption in fatty acid synthesis while increasing NADPH generation via fatty acid oxidation.⁵⁵ Additionally, AMPKα-dependent activation of the Nrf2 and AKT/FOXO3A pathways upregulated the expression of the antioxidant genes HO-1 and NQO-1, along with glutathione biosynthesis, providing an extra layer of antioxidant protection. The inactivation of both FOXO1/3A proteins and Gsk3β via PI3K/AKT activation closely correlates with AMPK activation, further supporting its critical role in the mechanisms of CP2 and C458.⁵⁶ In AD, inactivation of GSK3β offers significant neuroprotection by reducing Tau hyperphosphorylation and Aβ production,⁵⁷ whereas inactivation of FOXO1/3A prevents neuronal apoptosis, improving insulin signalling and metabolic function.⁵⁸ Downregulation of GSK3β, FOXO1A and FOXO3A activity has been shown to reduce inflammation and synaptic dysfunction.^57,58 Thus, targeting diverse pathological processes in AD can be achieved by directly modulating the adaptive stress response via weak inhibition of mtCI with the small molecules CP2 or C458. This approach coordinates multiple pathways to restore mitochondrial function, maintain energy homoeostasis, and ultimately provide neuroprotection. Notably, C458-related neuroprotective mechanisms critically depend on activation of AMPKα. In AMPKα1/α2 knockout MEFs, C458 treatment fails to activate the adaptive stress response, exacerbating mitochondrial dysfunction, oxidative stress, and inflammation while preventing cellular adaptation to mild energetic stress.

Lower ATP levels in AD brains reflect underlying mitochondrial dysfunction and may increase neuronal sensitivity to mtCI inhibition. This condition likely enhances the efficacy of mild mtCI inhibitors such as C458 by promoting AMPKα2-mediated neuroprotective signalling. However, it also highlights the importance of precise dosing to avoid excessive bioenergetic stress. Our data demonstrate that C458 activates adaptive stress pathways, including PGC1a, ULK-1, HO-1, and IκBα, and elevates NADPH and GSH levels at nanomolar concentrations within the EC₅₀ range, indicating a favourable therapeutic window even in metabolically compromised neurons.

Support for the safety of pharmacological mtCI inhibition in ageing populations includes the use of metformin, a mild and non-specific mtCI inhibitor, and resveratrol, which also targets mtCI. Notably, metformin and resveratrol share key mechanistic features with CP2 and C458, including mild mtCI inhibition that activates AMPK, induces antioxidant defences, and improves mitochondrial bioenergetics. These convergent mechanisms contribute to the neuroprotective and health-promoting effects observed across multiple models of neurodegeneration and ageing.

Activation of AMPKα in response to mild mtCI inhibition and changes in the AMP/ATP ratio,¹³ rather than direct activation of AMPK, appears to be a promising strategy for neuroprotection. The diversity of AMPKα heterotrimeric complexes and their tissue-specific distribution present a challenge in targeting AMPKα activation specifically in the brain or neurons without causing potential adverse effects in other tissues.⁵⁹ Direct AMPK activators act systemically and broadly, bypassing the AMP/ATP-sensing mechanism and activating AMPK without causing true energetic stress. Systemic and chronic AMPK activation has been associated with adverse metabolic consequences.^60,61 In contrast, mtCI inhibitors such as C458 offer a more targeted approach, preferentially activating AMPK in mitochondria-rich tissues, such as the brain,¹³ suggesting that these inhibitors promote mitochondrial adaptation in select organs without significantly activating AMPK in tissues that are less dependent on OXPHOS-driven metabolism. Additionally, some mtCI inhibitors, such as metformin⁶² and MitoTam,⁶³ exhibit specificity because they accumulate in the mitochondrial matrix, further enhancing their targeted action on mitochondria. While we did not test whether C458 accumulates in mitochondria, we cannot rule out this possibility because of its similarity with the mechanism of CP2 action.¹³ Although C458 competes with rotenone for binding to the Qd site of mtCI, it only mildly inhibits mtCI, as evidenced by modest NADH accumulation and a gradual decline in the OCR. The neuroprotective action of C458 is blocked by nanomolar concentrations of rotenone, which irreversibly bind to the Qd site of mtCI, suggesting that C458 binds to the Qd site reversibly with low affinity. Interestingly, rotenone at picomolar concentrations reversed age-related gene expression changes, rejuvenated the transcriptome, and extended the lifespan of Nothobranchius furzeri.⁶⁴ These findings indicate that low doses of mtCI inhibitors, or weak inhibition, can activate the beneficial stress response. However, complete inhibition of mtCI by higher doses of rotenone induces ATP depletion, severe mitochondrial dysfunction, excessive ROS production, and oxidative stress, ultimately causing neurodegeneration. In contrast, C458 promotes mitochondrial adaptation without generating ROS, restoring energy balance and leading to neuroprotection.

Since AD is a multifactorial disease, various combination therapies⁶⁵ and multitarget⁶⁶ and dual-action⁶⁷ small molecules have been tested to increase treatment efficacy. While these therapies can address several pathological mechanisms, such as Aβ accumulation and inflammation, activation of the adaptive stress response via mtCI inhibition offers a more specific and targeted therapeutic approach, which may lead to fewer side effects, greater precision, and strong efficacy in AD, where mitochondrial dysfunction is one of the earliest symptoms and a contributing factor to disease development and progression.^4,68 Given this, mtCI inhibitors could be administered as a single drug or in combination with other medications relatively early in AD progression when mitochondrial function has not yet been severely impaired. This would reduce concerns regarding the additive effect of mtCI inhibition, which could exacerbate mitochondrial dysfunction. However, it is important to recognise that AD encompasses multiple subtypes with distinct molecular signatures, including varying degrees of mitochondrial dysfunction. Our data generated to date, using human cells and multiple mouse models of AD, mitochondrial disease (Ndufs4^−/−), and ageing, indicate that the mtCI inhibitors CP2 and C458 are both safe and effective, regardless of the baseline severity of mitochondrial impairment. Nevertheless, this paper describes a preclinical proof-of-concept study and broader validation across well-characterised AD subtypes will be essential for advancing mtCI inhibitors as precision therapeutics, and this remains a key focus of our future studies. Together with evidence supporting the strong safety profile of metformin in ageing populations, and our own data demonstrating neuroprotection following CP2 administration in APP/PS1 mice with reduced mitochondrial function,^14,69 our findings in Ndufs4^−/− mice show that CP2 enhances mitochondrial biogenesis and turnover without inducing toxicity or affecting lifespan, even when mtCI activity is reduced by approximately 50%.⁷⁰

The dual effect of reducing p-Tau and the Aβ42/Aβ40 ratio in APOE4-expressing human organoids makes C458 a particularly attractive compound for addressing major hallmarks of AD in a single therapeutic. This multitarget action is especially significant in AD, where amyloid and p-Tau pathologies are interlinked and often synergistically contribute to neurodegeneration. The ability of C458 to reduce both p-Tau and Aβ levels could slow or prevent early neurodegenerative processes in APOE4 carriers. This is critical, as APOE4-related neurodegeneration tends to occur early in life, and delaying these pathologies could significantly impact cognitive health. Given its ability to alleviate inflammation, increase antioxidant capacity, increase mitochondrial bioenergetics and autophagy, and promote neuroprotective responses, this study highlights the potential of mtCI inhibitors as treatments for age-related neurodegenerative diseases.

While the present study focuses on preclinical models, we acknowledge that species-specific differences in drug absorption, distribution, metabolism, and excretion (ADME) between mice and humans may affect translational outcomes. To address this, future studies will include in vitro analyses using human liver microsomes and hepatocytes to evaluate the metabolic stability of mtCI inhibitors, identify primary metabolic pathways, and characterise the involvement of cytochrome P450 enzymes. These data will support the development of physiologically based pharmacokinetic (PBPK) models to predict drug disposition and elimination in humans. In addition, given that individuals with AD are commonly treated with multiple medications, evaluating the potential for drug-drug interactions will be a critical next step. Collectively, these studies will inform clinical translation and support the safe use of C458 and other mtCI inhibitors in clinical settings.

The study's strengths include a development of a robust drug-discovery funnel that prioritises safe, efficacious small-molecule mtCI inhibitors using human cells and AD mouse models; replicated efficacy testing in two independent mouse cohorts across diverse functional endpoints; use of translational outcomes aligned with future clinical trials; rigorous target engagement and mechanism-of-action confirmation via complementary assays and genetic/pharmacological perturbations; and validation in human AD organoids. A key limitation is the absence of sex-specific analyses, which may be important for personalised medicine. Additional studies are underway to address the remaining questions.

Contributors

E.T. conceived the study, assembled the multidisciplinary team of collaborators, and received funding for the project. S.T., A.S., S.I.M., T.K.O.N., M.O., L.Z., and J.T.D. performed the experiments and analysed and interpreted the data. A.S. and S.Y.C. conducted the electrophysiology experiments. T.K.O.N., T.N., W.L., and T.K. conducted the experiments on the human organoids, and S.T. and E.T. wrote the manuscript. All authors read and approved the final version of the manuscript. E.T. and S.T. directly assessed and verified the underlying data presented in the manuscript.

Data sharing statement

Source data is provided with this paper. The C458 structure was deposited in PubChem with CID 134259587. All data regarding the experiments described in the paper are available upon request to the corresponding author Eugenia Trushina (trushina.eugenia@mayo.edu). C458 and related compounds have IP protection in US Patent # 10,336,700, July 2, 2019; US Patent # 10,774,045, September 15, 2020; US Patent # 11,161,814, February 11, 2021; US Patent # 12,017,992 June 25, 2024.

The manuscript was professionally edited by Rubriq. No portion was generated using AI.

Declaration of interests

Dr. Trushina is a coauthor of four U.S. Patents, 11,161,814; 10,336,700; 10,774,045; and 12,017,992, relevant to the development of the compound described in the paper. She and the Mayo Clinic own the IP on this technology. The authors declare that they have no conflicts of interest.