Abstract
Parkinson’s disease (PD) involves multiple pathological processes in midbrain dopaminergic (mDA) neurons, including protein degradation defects, vesicular trafficking disruption, endolysosomal dysfunction, mitochondrial issues, and oxidative stress. Current PD models often lack complexity and focus on single phenotypes. We used patient-derived SNCA triplication (SNCA-4x) and isogenic control (SNCA-corr) mDA neurons, applying high-content imaging-based morphological profiling to identify and rescue multiple phenotypes. Screening 1,020 compounds, we identified top-scoring compounds that restored healthy profiles in SNCA-4x neurons, increasing Tyrosine hydroxylase (TH) and decreasing α-synuclein (αSyn) levels. Several hits were linked to mitochondrial biology. Tyrphostin A9, a mitochondrial uncoupler, and several of its structural analogues decreased ROS levels, normalized mitochondrial membrane potential, and increased respiration. Western blotting confirmed that Tyrphostin A9 reduces αSyn levels. Our study highlights the neuroprotective potential of mild mitochondrial uncoupling in mDA neurons.
Similar content being viewed by others
Introduction
Parkinson’s disease (PD), with its diverse phenotypic, neuropathological, and genotypic manifestations, is increasingly acknowledged as a complex disorder rather than a singular disease entity1,2,3. While traditional models have been instrumental in devising symptomatic therapies to address PD’s motor symptoms, they have been less successful in developing neuroprotective strategies4. Genetic stratification of patients into subgroups has emerged as a crucial approach5,6. Over the past 25 years, significant progress has been made in identifying genes associated with PD, including those responsible for rare monogenic forms of the disease2. These monogenic forms have proven invaluable for PD research, as patient-based cell models exhibit disease-specific cellular phenotypes that mirror those found in postmortem brain tissue7,8,9,10,11. This approach suggests that targets for pharmaceutical interventions might be identifiable based on rare but potent molecular signatures12,13,14.
Point mutations, as well as duplications and triplications of the wild-type alpha-synuclein gene (SNCA), are responsible for causing autosomal dominantly inherited PD15,16,17,18. Patients with SNCA triplication (SNCA-4x) express α-synuclein protein (αSyn) from four alleles resulting in a roughly 2-fold increase of αSyn, and typically exhibit an early disease onset, accelerated disease progression, and significant dementia. In contrast, patients with SNCA duplications often display a more conventional late-onset PD phenotype. The severity of clinical trajectories in SNCA multiplication appears to correlate with gene dosage. Patient-derived induced pluripotent stem cells (iPSC) carrying the SNCA triplication have been successfully differentiated into midbrain dopaminergic (mDA) neurons and organoids19,20,21,22. Despite their young age, iPSC mDA neurons exhibit PD-relevant phenotypes. SNCA gene triplication leads to an increase in αSyn levels resulting in oligomeric αSyn pathology and elevated αSyn extracellular release19,21,23,24. Multiple pathophysiological defects correlate with increased αSyn protein expression, such as altered lysosomal and mitochondrial functions, dysregulated autophagy increased endoplasmic reticulum stress, and heightened oxidative stress levels20,2224–28. Due to the interconnection of many of the identified phenotypes, it is currently an open question whether a rescue of one key or multiple cellular PD hallmarks is necessary to achieve a significant neuroprotective effect. Imaging-based profiling is a strategy that transforms the rich data in biological images into multidimensional profiles of extracted features, which can be analyzed to reveal biological signatures, aiding in disease understanding and drug discovery, with advancements in machine learning and computational technologies enhancing its potential29. By capturing the observable characteristics of cells under different conditions, morphological profiling can provide valuable insights into the complex cellular mechanisms underlying PD8,9,11.
In this study, we find that SNCA gene triplication results in a multidimensional alteration of the morphological profile in mDA neurons. We further identify multiple small molecular compounds correcting this altered profile to align with the phenotype of isogenic controls (SNCA-corr) mDA neurons. One of the identified compounds, Tyrphostin A9, exerts its neuroprotective effect via mitochondrial uncoupling, which in turn reduces reactive oxygen species (ROS) and high molecular weight αSyn species. In summary, we show that screening small molecules in a PD-relevant human model system using a multi-phenotypical readout can identify modulators of not just a single, but multiple disease relevant phenotypes.
Results
SNCA-4x mDA neurons have a distinct but modulable morphological profile
We differentiated patient-derived SNCA gene triplication and isogenic control iPSCs into mDA neurons (SNCA-4x and SNCA-corr) for 30–37 days. These neurons showed the expected mDA neuron morphology and no de novo copy number variations. Both cell lines expressed neuro-typical markers β-tubulin III (90–94%), MAP2 (87–89%), and the dopaminergic neuron markers Tyrosine hydroxylase (TH) (84–88%) and FOXA2 (84–90% in TH positive cells) (Figure S1A-C). Western blotting revealed increased αSyn protein level and signs of αSyn aggregation in SNCA-4x mDA neurons after D70 (Figure S1D). A JC-1 dye assay indicated reduced mitochondrial membrane potential in D35 SNCA-4x mDA neurons (Figure S1E). For morphological profiling, mDA neurons were cultured until D36 in 384-well plates and stained with Hoechst, and antibodies against TH, αSyn, and MAP2 (Fig. 1A). Automated segmentation and feature quantification were used to overcome challenges in assessing subtle phenotypes. Image segmentation was performed on 16 fields per well and 127 morphological features were determined and expressed as means per well. Features were calculated for the whole cell, nucleus, cytoplasm or membrane and describe different aspects of the images (Fig. 1B, Table S4). Texture features capture the spatial distribution and arrangement of pixel intensities in an image. Context features provide information about the spatial relationship between signals, for example the correlation between image channels. Intensity features measure the fluorescence intensity and can provide insights into the expression levels of proteins. Count and area/shape features quantify the number, size and morphology of the cellular components. First, we examined basic morphological features of SNCA-4x and SNCA-corr mDA neurons. SNCA-4x neurons had fewer live cells and approximately 50% increased αSyn protein expression. In our study, we utilized Prostratin as a proof-of-concept molecule due to its proposed activity in degrading αSyn. Previous research has demonstrated that specific phorbol esters, such as PEP005 or Prostratin, can reduce αSyn through proteasomal degradation7. Based on these findings, we hypothesized that Prostratin’s activity would lead to an altered morphological profile in our experimental model. Treatment with Prostratin indeed reduced dead cells and αSyn expression in SNCA-4x neurons to levels similar to SNCA-corr neurons (Fig. 1C). Neuronal morphological profiles before and after compound treatment were created by scaling feature values and calculating mean values across wells (Fig. 1D). We used the mean absolute difference to find the features that differed the most between profiles. Comparing SNCA-corr and SNCA-4x neurons, the radial αSyn intensity profile, which measures fluorescence intensity as a function of distance from a central point, differed most. Prostratin-treated and untreated SNCA-4x neurons also varied in number of nuclei, total MAP2 cellular surface, and αSyn staining contrast (Fig. 1E). These findings show that small molecular treatment can rescue morphological features in mDA neurons which can be measured by morphological profiling.
SNCA-4x mDA neurons have a distinct but modulable morphological profile. (A) SNCA-4x and SNCA-corr mDA neurons were differentiated until D33, treated with 5 µM Prostratin or DMSO for 72 h and immunofluorescently stained. Image channels were segmented for morphological feature quantification. (B) Schematic overview of image data generation and morphological feature types. (C) Raw morphological feature values in control, and untreated and Prostratin treated SNCA-4x mDA neurons. Each datapoint represents technical replicate data from one 384-well plate well. Larger markers represent the medians each independent experiments. The boxplots show the median and the 2nd and 3rd quartiles of the data. Whiskers indicate 1.5X of the interquartile range. Welch’s unequal variances t-test was performed using the medians of independent differentiation experiments. ns = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. (D) Scaled well and replicate median feature values aggregated into morphological profiles. Cosine similarity was used for clustering. (E) Top ten absolute mean differences between features of SNCA-4x and Control or SNCA-4x neurons treated with Prostratin. Means with standard deviations are shown. (F) Two first principal components (PCs) explaining most variation in the dataset after Principal Component Analysis (PCA). Each data point represents mean data from one well. (G) Absolute Pearson correlation of all features with the first ten PCs after PCA. Representative images illustrate cellular αSyn staining patterns. All data in this figure was generated in four independent experiments each containing at least 12 technical replicates per condition.
Next, we performed Principal Component Analysis (PCA) using the morphological profile data to visualize overall differences between compound-treated and untreated cells. The first two principal components (PCs) accounted for 57% of the variance and separated the tested cell lines and Prostratin-treated neurons, indicating significant differences between them (Fig. 1F). The remaining PCs each accounted for 11% or less of the variance, suggesting lesser contribution to the overall variance (Figure S1F). To understand which features were mainly responsible for distinguishing the experimental categories, we calculated the absolute Pearson’s correlation between the first ten PCs and each morphological feature. PC1 correlated mostly with αSyn signal texture-related features, indicating a different αSyn subcellular distribution pattern in the neurons. PC2 and PC3 correlated with signal radial distribution and nucleus-related features, similar to the results we obtained when comparing experimental conditions pairwise using the mean absolute difference (Fig. 1G, Figure S1G). Manual inspection of images confirmed these findings, showing grainy αSyn localization outside the nucleus in SNCA-4x cells (Fig. 1G). This indicated that radial distribution and texture features indeed capture αSyn localization patterns. As a result, high-dimensional morphological profiling, by capturing a wide array of cellular characteristics, allowed us to identify subtle differences between experimental conditions that may not be apparent through visual inspection alone.
Compound modulation shifts a SNCA-4x towards a control morphological profile
Based on our observation that small molecular compounds, such as Prostratin, can partially rescue SNCA-4x mDA neuron morphological profiles, we wondered whether more potent small molecules could be identified using a morphological profile screening approach. We assembled a 1,020 compound library which included 734 FDA-approved compounds, 435 of which can pass the blood-brain-barrier, 120 molecules targeting PD-relevant processes, and 166 diverse bioactive molecules (Fig. 2A, Table S3). Cryopreserved D30 mDA neurons were cultured in 384-well plates until D33, treated with 5 µM of each compound for 72 h and stained (Fig. 2B). Each compound was tested in 4 replicates across 20 384-well plates, with 36 control wells per plate. Control conditions established baseline and maximum effect profiles (SNCA-4x + DMSO and SNCA-corr + DMSO, respectively). Morphological profiles were generated from 307,200 images and all data was aggregated per well (Fig. 2B).
Compound modulation shifts a SNCA-4x towards a control morphological profile. (A) Overview about screening library composition and neuron treatment scheme. (B) Schematic overview of image data generation. Morphological features were calculated from imaging data to measure compound-induced morphological profile changes from a SNCA-4x towards a SNCA-corr-like profile. (C) Schematic of morphological profile data processing workflow. (D) Overview of morphological feature types used for hit selection after removal of redundant features using a Pearson correlation threshold of 0.95. (E) Morphological profile Mahalanobis distances of compound treated mDA neurons relative to SNCA-4x and SNCA-corr morphological profiles. The black dashed line indicates the selection window based on 4σ from the SNCA-4x median. Each datapoint represents mean data from one well and for compounds the mean value of all four replicates is shown. (F) Compounds inducing morphological profiles with the 10 shortest mean Mahalanobis distances to the median SNCA-corr morphological profile. (G) Selected hits clustered by the Cosine similarity of their respective morphological profiles. Compounds are color coded by attributed biological activity. (H) Tanimoto similarities of top 10 selected compounds.
We aimed to identify compounds shifting SNCA-4x profiles towards SNCA-corr-like phenotypes. Morphological profiles were Z-score normalized for comparability (Fig. 2C, Figure S2A). Pearson’s correlation reduced the initial 127 features to 40 non-redundant ones, mostly describing cellular area, shape, and TH and αSyn staining intensities (Fig. 2D, Figure S2B). Plate quality was assessed using the Strictly Standardized Mean Difference (SSMD) of the Mahalanobis distances between treatment groups. All plates showed significant differences (Figure S2C). Linear Discriminant Analysis (LDA) was used to compare each compound profile with the reference profiles. LDA achieved a classification accuracy F1 score of 0.96 in a test data set and 0.93 ± 0.04 in a 10-fold cross-validation, indicating a high ability to correctly predict whether compound-induced morphological profiles resemble one of the reference classes (Figure S2D). Most compounds had no effect on the SNCA-4x profile, while some induced SNCA-corr-like profiles (Figure S2E).
We aimed to identify compounds that could shift SNCA-4x profiles towards SNCA-corr-like phenotypes. Morphological profiles were compared using the Mahalanobis distance, a metric that measures how different two profile vectors are. We identified 28 compounds with a mean distance of at least 4 to the SNCA-4x median profile (Fig. 2E, Table S1). Of these, the 10 compounds with the shortest distance to the SNCA-corr profile were considered to partially rescue the phenotype. Tyrphostin A9 and Alectinib were the most similar to the SNCA-corr profile (Fig. 2F). To validate our findings, we used Cosine similarity, a metric that measures whether two vectors (or morphological profiles) are pointing in the same direction. Nine out of the 10 compounds with the shortest Mahalanobis distance to SNCA-corr grouped in the same cluster as SNCA-corr, supporting a distance and directional similarity between their morphological profiles (Fig. 2F and G). Although the top hits’ morphological profiles were highly similar among each other and compared to the SNCA-corr profile, their chemical structure-based Tanimoto similarities differed considerably (Fig. 2H). We validated the effect of the identified compounds on cellular phenotypes using data from the Joint Undertaking in Morphological Profiling (JUMP) consortium, an effort to create compound-induced morphological profiles in U2OS cells by CellPainting30,31. Also here, in a different cell type and using a different staining assay, we observed that compounds with similar mode of action induced similar morphological profiles (Figure S2F). This supports the notion that the identified compounds, while acting on different targets, might possibly acting on related pathways leading to a similar phenotypic outcome. Tyrphostin A9 was originally developed as an epidermal growth factor receptor (EGFR) inhibitor, but similar to other Tyrphostins has been linked to mitochondrial membrane fragmentation, reduced OXPHOS, and potentially neuroprotection via reduction of ROS32,33,34,35. Since mitochondrial defects play an important role in PD pathology, we decided to further characterize Tyrphostin A9’s molecular mechanism in mDA neurons3.
Tyrphostin A9 rescues the overall mDA neuron morphological profile in a dose dependent manner
To establish hit compound dose-response effects we focused on cell viability and αSyn level assessment as well as on morphological profiling using six different compound concentrations ranging from 0.25 to 20 µM and a treatment duration of 72 h (Figure S3, Figure S4). For Tyrphostin A9, we observed that higher compound concentrations in SNCA-4x neurons resulted in a higher resemblance to SNCA-corr neuron morphological features (Fig. 3A). To quantify this observation, we calculated the Cosine distance of the compound treated SNCA-4x morphological profile to the SNCA-4x + DMSO and SNCA-corr morphological profiles as a function of compound concentration. Cosine distance can be defined as 1 – Cosine similarity. The intuition behind this is that if two vectors are perfectly the same then the similarity is 1 (angle = 0) and thus, the distance is 0 (1–1 = 0). We found that increasing compound concentrations decreased the distance between SNCA-4x and SNCA-corr morphological profiles, therefore increasing their similarity (Fig. 3B). The opposite is true when SNCA-4x compound treated neurons were compared to SNCA-4x DMSO-treated morphological profiles. Here, the Cosine distance increased with increasing compound concentration, reflecting the change of phenotype towards the SNCA-corr profile. Next, we investigated the effect of increasing compound concentrations on three single morphological features: αSyn intensity, TH intensity and MAP2 neurite length (Fig. 3C). Tyrphostin A9 concentrations > 1 µM decreased the nuclear and cytoplasmic αSyn staining intensity. Tyrphostin A9 also positively impacted TH staining intensity. At higher concentrations we observed little effects on MAP2 neurite length, indicating no acute toxicity.
Tyrphostin A9 rescues the overall morphological profile in a dose dependent manner. (A) Line plot representation of morphological profiles of Tyrphostin A9 treated and untreated mDA neurons. D33 neurons were treated with 0.25-20 µM compound for 72 h before fixation. All used 56 features are listed in Table S5. (B) Cosine distances of SNCA-4x neurons as a function of compound concentration relative to DMSO-treated SNCA-4x and SNCA-corr neurons. (C) Dose-response curves for the three selected morphological features αSyn intensity, TH intensity, and MAP neurite length. Representative images of the same cell in each fluorescent channel are shown for DMSO and 5 µM compound treatment. Each compound concentration was tested at least in triplicate. Mean and 95% confidence interval are shown.
Multiple identified hits induce mitochondrial uncoupling and ROS reduction
Tyrphostin A9 has been explored in vitro for its roles as neuroprotective agent, possibly functioning as an oxidative phosphorylation uncoupler reducing ROS32,33,36. Mitochondrial uncoupling involves the decrease of the mitochondrial proton gradient disrupting the coupling between the electron transport chain and ATP synthesis. Whether Tyrphostin A9 can exert these effects in mDA neurons is currently unclear. We first assessed Tyrphostin A9’s ability to modulate the mitochondrial membrane potential using the JC-1 dye. We treated D33 mDA neurons for 72 h with increasing concentrations of Tyrphostin A9 (0.01 µM to 10 µM) and observed a decrease in red fluorescence for concentrations > 0.5 µM, indicating a concentration dependent decrease in mitochondrial membrane potential (Fig. 4A). Valinomycin was used as positive control inducing a very low membrane potential. While FCCP and Tyrphostin A9 lowered the mitochondrial membrane potential at the lowest doses, also other molecules with morphological profiles similar to Tyrphostin A9, such as Tinostamustine or VPS34 inhibitor 1, led to a decreased mitochondrial membrane potential (Figure S5). Since a reduction in mitochondrial membrane potential is often associated with decreased ROS levels, we used a dye-based ROS assay to measure ROS levels in mDA neurons upon compound exposure. Tyrphostin A9 reduced ROS in a dose-dependent manner. At 2.5 µM Tyrphostin A9 reduced ROS down to the level of the positive control antioxidant N-acetyl-cysteine (NAC) (Fig. 4B). Interestingly, several other molecules that were pre-selected for inducing morphological profiles similar to the SNCA-corr phenotype, such as Simvastatin, Trifluoperazine, or WAY-600, also lowered neuronal ROS levels (Fig. 2G, Figure S6). We hypothesized that Tyrphostin A9 might lower mitochondrial membrane potential and ROS due to its function as protonophore facilitating proton transfer across membranes. We used a Rhod-FF assay which measures intracellular Ca2+ levels and found only small baseline differences between SNCA-corr and SNCA-4x mDA neurons, while the addition of Tyrphostin A9 to SNCA-4x mDA neurons increased intracellular Ca2+ levels by more than 70% (Fig. 4C). Addition of the known protonophore Ionomycin further increased intracellular Ca2+ levels demonstrating that Tyrphostin A9 likely uncouples the mitochondrial membrane potential via facilitating the entry of protons, thereby decreasing oxidative phosphorylation and ROS levels. In neurons protonophores are also expected to disrupt the neuronal membrane potential by decreasing the proton gradient, leading to a decrease in the driving force for the generation of action potentials, which might negatively impact neuronal firing. Using a microelectrode array (MEA) we measured the neuronal firing rate before and after Tyrphostin A9 treatment (Fig. 4D). Before treatment ca. 60-day old SNCA-corr mDA neurons showed approximately 2.5-fold higher firing rates than SNCA-4x neurons. When treating the mDA neurons for up to 72 h with 2.5 µM Tyrphostin A9 we observed a sharp drop in firing already 30 min after treatment. After > 24 h of treatment a slight recovery effect was observed and firing rates began to increase mildly. Taken together, our findings suggest that the identified compounds, although affecting different targets, might be influencing related pathways, resulting in a similar phenotypic outcome such as mitochondrial membrane potential and ROS lowering.
Tyrphostin A9 acts as ROS-reducing mitochondrial uncoupler in mDA neurons. (A) mDA neurons were treated with 0.1–10 µM Tyrphostin A9 for 72 h and the mitochondrial membrane potential was measured using the JC1 dye. JC1 fluoresces red at high membrane potential and green at lower membrane potentials. Three independent experiments with at least three technical replicates each were performed. (B) mDA neurons were treated with 0.5, 1, and 2.5 µM Tyrphostin A9 for 72 h and the ROS level was determined using the CellROX dye. The obtained signal was normalized to total protein content. Three independent experiments with at least three technical replicates each were performed. (C) Left panel: Scheme illustrating the protonophore mode of action. Right panel: mDA neurons were treated with 0.5, 1, and 2.5 µM Tyrphostin A9 for 72 h and the intracellular Calcium level was measured using the Rhod-FF dye for 250 s before and after the addition of Ionomycin. The obtained signal was normalized to total protein content. Two independent experiments with at least three technical replicates each were performed. (D) Neuronal activity was measured in ca. 60-day old mDA neurons using a microelectrode array (MEA). The neuronal firing rate was assessed before and up to 72 h after treatment with 2.5 µM Tyrphostin A9. Neuronal activity was measured in two independent experiments with at least three technical replicates each. For all data, the mean and 95% confidence interval is shown.
The tyrphostin A9 structural analogue AG879 induces mitochondrial uncoupling
Tyrphostins represent a large family of small molecules with various attributed neuroprotective modes of action. Next to their role as mitochondrial uncouplers, Tyrphostins have been described as antioxidants or as having a positive effect on cellular glutathione levels33. We therefore compared a range of Tyrphostin A9 structural analogues using various biochemical assays to establish a Tyrphostin A9 structure-function relationship in mDA neurons (Fig. 5A). FCCP is a potent mitochondrial uncoupler used during oxygen consumption rate (OCR) measurements and facilitates proton transport across mitochondrial membranes. This action collapses the proton gradient and disrupts the mitochondrial membrane potential. As a result, the electron flow through the Electron Transport Chain (ETC) is uninhibited, allowing oxygen consumption by complex IV to reach its maximum. The resulting OCR increase is used as an estimation of the maximal respiratory capacity of the cell (Fig. 5B, left panel). Next to its use as a control compound, we also detected FCCP as one of the top hits during morphological profiling (Fig. 2F). We therefore compared Tyrphostin A9 and its structural analogues to FCCP using an OCR assay. We found that, similar to FCCP, Tyrphostin A9 and its closest structural analogues AG1024 and AG879 led to a strong increase in maximal respiratory capacity, highlighting their mitochondrial uncoupling activity (Fig. 5B, right panel). Next, we determined compound cytotoxicity by measuring Lactate dehydrogenase (LDH) in the culture medium. LDH is a stable enzyme, present in all cell types and rapidly released into the culture medium upon damage of the plasma membrane. None of the tested compounds induced significantly more cell death than the DMSO control when used at 0.5–2.5 µM for 72 h. Cell lysate was used as control corresponding to the maximum possible amount of released LDH (Fig. 5C). For Tyrphostin A9, we also tested concentrations ranging from 0.625 to 10 µM for 1–5 days and did not measure any increase of released LDH corresponding to low cytotoxicity (Figure S9). JC1 dye fluorescence was measured to determine the mitochondrial membrane potential. Tyrphostin A9 and its most closely related structural analogue Tyrphostin AG879 decreased the membrane potential as measured by the 547 nm/488nm fluorescence ratio from 0.95 to approximately 0.2. The second most similar analogue Tyrphostin AG1024 also decreased the membrane potential, although to a lesser extent (Fig. 5D). In line with the decreased membrane potential Tyrphostins A9 and AG879 also reduced ROS levels to ca. 80% of the DMSO baseline level (Fig. 5E). Next, we assessed glutathione levels. In healthy cells, both reduced glutathione (GSH) and its oxidized form (GSSG) are present, with the reduced form (GSH) making up the majority, while the oxidized form (GSSG) increases during oxidative stress. Although we expected an GSH increase due to ROS reduction, we observed that Tyrphostins A9, AG879 and AG1024 decreased GSH ratio, possibly due to lower overall glutathione synthesis after ROS decrease. Tyrphostin A9 lead to the strongest decrease of 48% at 2.5 µM (Fig. 5F). Mitochondrial uncoupling also strongly reduces cellular ATP production. We therefore tested the effects of all compounds on mDA neuron ATP level. We found that the compounds with the strongest uncoupling effects (Tyrphostins A9, AG1024, and AG879) reduced ATP levels the most (Figure S7A). mDA neuron treatment with the mitochondrial complex I inhibitor Rotenone led to similarly low ATP levels as Tyrphostin A9 treatment alone, indicating a near maximal Tyrphostin A9 uncoupling effect (Figure S7B). Since close Tyrphostin A9 structural analogues such as AG879 and AG1024 reproduce its cellular effects, the shared structural features may be critical for the molecule’s function in mDA neurons.
Tyrphostin A9 structural analogues reduce mitochondrial membrane potential and ROS. (A) Molecular structures of Tyrphostin A9 and five selected structural analogues including their names, Compound ID (CID), and Tanimoto similarity coefficient relative to Tyrphostin A9. (B) Oxygen consumption rate (OCR), a proxy readout for mitochondrial respiration, was measured in D30 mDA neurons treated with 2.5 µM compound. (C-F) D33 mDA neurons were treated with 0.5–2.5 µM of each compound for 72 h. (C) Lactate dehydrogenase (LDH) activity was measured in the culture medium to determine relative cytotoxicity compared to cellular lysate (= 100%). (D) Mitochondrial membrane potential was measured using the JC-1 dye. (E) ROS level was determined using the ROS-ID detection kit and the obtained signal was normalized to total protein content. (F) Reduced glutathione (GSH) was measured using monochlorobimane (MCB) and the obtained signal was normalized to total protein content. Three independent experiments with at least three technical replicates were performed. The mean and 95% confidence interval are shown. Dashed lines indicate the median values of the respective controls.
Tyrphostin A9 induces αSyn Lowering
Our phenotypic compound screen revealed that mDA neuron αSyn levels were lowered by Tyrphostin A9 treatment (Fig. 3C). Since immunofluorescent staining is not sensitive enough to differentiate between αSyn monomers, oligomers, and larger structures such as fibrils, we performed Western blotting of Tyrphostin A9 treated and untreated > D50 mDA neuron lysates (Fig. 6A). We observed a 2-fold reduction of both low molecular weight αSyn (≈ 15 kDa) as well as higher molecular weight αSyn species. We observed a similar Tyrphostin A9-induced decrease of αSyn in an additional genetic Parkinson’s disease model. mDA neurons carrying a Miro1 R272Q mutation showed a 70–80% decrease of 15 kDa and 33 kDa αSyn species (Figure S10)37. Most cellular proteins are selectively degraded by the proteasome via ubiquitination. Caspase and chymotrypsin are part of the proteasomal protein degradation machinery, contributing to the proteolytic activities of the proteasome. Using plate reader assays we therefore checked how caspase and chymotrypsin activities are impacted by Tyrphostin A9 in mDA neurons. Both enzymes’ activities were not significantly different between SNCA-corr and SNCA-4x mDA neurons, while addition of Tyrphostin A9 rather decreased (≈ 50%) or did not impact caspase- or chymotrypsin-related proteasomal degradation, respectively (Fig. 6B). As expected, the known proteasome inhibitor MG132 strongly decreased both caspase and chymotrypsin activities in mDA neurons by > 70%.
Tyrphostin A9 decreases αSyn protein expression. (A) Anti-αSyn Western blotting of mDA neuron lysates. Before lysis > D50 mDA neurons were treated with 0.3 or 2.5 µM Tyrphostin A9 for 72 h. Three independent neuronal differentiations were performed. Gel images are cropped. Full-length blots/gels are presented in Figure S8. (B) Caspase and Chemtrypsin activities were determined in approximately D30 mDA neurons treated with 1 or 2.5 µM Tyrphostin A9 or the proteasome inhibitor MG132 for 72 h. Three independent experiments with at least three technical replicates each were performed. The boxplots show the median and the 2nd and 3rd quartiles of the data. Whiskers indicate 1.5X of the interquartile range. (C) SNCA-4x mDA neurons were treated with 2.5 µM Tyrphostin A9 for 72 h and Western blotting was performed using antibodies against p-AMPK and β-actin. (D) SNCA-4x mDA neurons were treated with 2.5 µM Tyrphostin A9 for 72 h and Western blotting was performed using antibodies against p62, LC3B, and β-actin. Bar plots indicate the means of two or three independent differentiations. Welch’s unequal variances t-test was performed using the (technical) replicates of all independent experiments. ns = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. Uncropped Western blot images are shown in Figure S8 and Figure S11. (E) Scheme depicting the use of morphological profiling to identify compounds that produce a healthy control-like (SNCA-corr) profile in mutant (SNCA-4x) mDA neurons. The identified compound Tyrphostin A9 increases mitochondrial uncoupling and decreases ROS and αSyn levels which can explain the rescued morphological profile.
Since Tyrphostin A9 did not positively impact proteasomal protein degradation in mDA neurons, the observed lower αSyn levels after compound treatment might result from a modulated alternative gene expression mechanism. To rule out transcriptional effects on α-synuclein (SNCA), we performed qPCR and found no significant change in SNCA mRNA levels following Tyrphostin A9 treatment (Figure S12). Next, we investigated the effect of Tyrphostin A9 on autophagy-related pathways in D38 mDA neurons. Cells were treated with 2.5 µM Tyrphostin A9 for 72 h, and protein levels were assessed by Western blotting. We observed increased phosphorylation of AMPK (p-AMPK) upon Tyrphostin A9 treatment, which is consistent with our findings that mitochondrial uncoupling by Tyrphostin A9 reduces cellular ATP levels, a known trigger for AMPK activation (Fig. 6C, Figure S7A). In line with this, we observed an increase in the levels of p62/SQSTM1, a selective autophagy receptor, which is continuously synthesized during autophagy initiation (Fig. 6D). Additionally, LC3B levels were elevated, and two distinct bands corresponding to LC3-I and LC3-II were detected, suggesting altered autophagosome dynamics (Fig. 6D). The conversion from LC3-I (cytosolic form) to LC3-II (lipidated, membrane-bound form) reflects autophagosome formation38.
Discussion
Here we used microscopy-based morphological profiling to characterize the effects of 1,020 FDA-approved or PD pathobiology-related small molecules on the overall morphological appearance of human iPSC-derived SNCA-4x mDA neurons. We observed that chemical compounds exist that impact two important hallmarks of PD pathophysiology: ROS reduction via mitochondrial uncoupling and αSyn reduction (Fig. 6C). We scored compounds based on their potential to induce a morphological profile resembling SNCA-corr mDA neurons (Fig. 2). Phenotypic or morphological profiling is increasingly used in iPSC-derived cells, primary cells or brain slice systems to study neuronal diversity and to complement single cell RNA sequencing data39,40,41,42. Previously, we established morphological profiling in LRRK2 G2019S mutated iPSC-derived mDA neurons and showed that multiple LRRK2 inhibitors can partially rescue the neuronal morphological profile8. Drug screens in mDA neuron model systems have been performed before, however either using only small screening libraries, non-human cell systems such as zebrafish, or very young mDA neuron progenitors (< 20 days of differentiation)43,44,45,46. This study combines detailed morphological profiling in human mDA neurons with a compound screen using a larger library (> 1,000 compounds). Although morphological profiling has the potential to identify completely novel targets, a limitation is that it typically provides little information about the actual potential compound targets, requiring detailed target and mode of action deconvolution studies.
In our study, we focused on mitochondrial uncoupling since two of our top scoring compounds (FCCP, Tyrphostin A9) are annotated as protonophores and mitochondrial uncouplers. Originally developed as an EGFR inhibitor, Tyrphostin A9 stood out due to its αSyn reducing and TH increasing effects in our morphological profiling assay as well as the molecule’s previously described role in neuroprotection (Fig. 3)33,34,35. We found that the molecule promotes mitochondrial uncoupling and reduces ROS in mDA neurons in line with its previously described protonophore properties in other cell types (Fig. 4)32,33,36. In mDA neurons we made the observation that next to its mitochondrial uncoupling effect, Tyrphostin A9 also decreased neuronal activity which relies on charge-dependent action potentials (Fig. 4D). Tyrphostin A9 did not permanently damage the neurons, indicated by the gradual partial activity recovery over time, intact morphology, and low extracellular LDH levels (Fig. 3, Fig. 4D, Fig. 5C, Figure S9). Although the various Tyrphostins we tested were expected to increase GSH levels by reducing ROS, they instead decreased the GSH ratio. This might be due to reduced glutathione synthesis following ROS reduction, highlighting that the observed GSH depletion may reflect adaptive metabolic reprogramming and activation of protective pathways like Nrf2 signaling, rather than indicating toxicity47,48,49.
A surprising observation we made is Tyrphostin A9’s ability to reduce the low and high molecular weight αSyn levels in SNCA-4x and Miro1 R272Q mDA neurons by up to 80% (Figs. 3C and 6, Figure S10). We excluded a direct effect on caspase or chymotrypsin related proteasomal degradation mechanisms since Tyrphostin A9 either did not impact or even decreased these modes of protein degradation. Except for one study describing a role of Tyrphostin AG1024 in accelerating the degradation of Retinoblastoma Protein (pRb) likely involving the proteasome, Tyrphostins have not been directly linked to enhanced protein degradation50. Next to proteasomal protein degradation, the involvement of lysosomes in Parkinson’s disease progression through two major autophagic pathways, chaperone-mediated autophagy and macroautophagy, and their central role in degrading αSyn is well documented2. In Tyrphostin A9 treated mDA neurons, we detected elevated levels of Thr172 phosoporylated AMPK, an activator of autophagy (Fig. 6C). AMPK typically becomes phosphorylated once ATP levels are low, an effect that we also attributed to Tyrphostin A9 due to its mitochondrial uncoupling activity (Figure S7A). We detected elevated levels of p62/SQSTM1, a selective autophagy receptor that is normally degraded during autophagic flux. LC3B levels were also increased, with Western blot analysis revealing two distinct bands corresponding to LC3-I and LC3-II, indicative of changes in autophagosome dynamics (Fig. 6D)38. While these findings are consistent with the initial activation phase of autophagy during which p62 transiently increases, the accumulation of p62 also raises the possibility of impaired autophagic clearance51,52,53. One study showed that Tyrphostin AG490, via JAK2 inhibition, prevented the hormone-receptor complex Epo-EpoR from targeting lysosomes in the UT-7 human cell line, highlighting a possible link between Tyrphostin action and lysosomal biology54. Taken together, the ROS-lowering effect of several Tyrphostins, likely due to mild mitochondrial uncoupling, may help prevent α-synuclein misfolding and facilitate its degradation through a lysosome-associated pathway. Autophagy might be further induced by Tyrphostins due to mitochondrial uncoupling, the resulting decreased ATP levels, and AMPK phosphorylation. However, the exact mechanism underlying the Tyrphostin-induced reduction in α-synuclein levels and associated morphological rescue remains to be clarified in future studies.
Tyrphostins have been described as mitochondrial uncouplers for almost half a century, while mitochondrial uncoupling as a neuroprotective strategy is more recent, but has faced several safety-related hurdles during pre-clinical drug development55,56,57,58. Recently the blocking of mitochondrial complex I in pro-inflammatory microglia has been shown to protect the central nervous system against neurotoxic damage in mice59. While our current study focuses on iPSC-derived neurons, future investigations into mitochondrial uncoupling as a neuroprotective strategy may benefit from exploring potential non-cell autonomous mechanisms. Such mechanisms could be more effectively studied using advanced iPSC-derived co-culture systems or organoids that incorporate multiple central nervous system cell types, which are increasingly enabling the molecular characterization of complex cellular interactions60,61,62. Importantly, we do not propose Tyrphostin A9 itself as a therapeutic candidate for Parkinson’s disease or other synucleinopathies. Rather, our study demonstrates that morphological profiling in a PD-relevant human model system can uncover compound classes with potentially underappreciated neuroprotective mechanisms. Tyrphostin A9 serves as a tool compound to explore such mechanisms, including mild mitochondrial uncoupling and enhanced lysosomal degradation of α-synuclein. Future research should investigate the long-term safety of mitochondrial uncoupling strategies, ideally in vivo, and further explore the link between Tyrphostin A9-induced lysosomal modulation and neuroprotection.
Materials and methods
Origin of iPSC lines and differentiation into mDA neurons
SNCA iPSC lines (originally named AST23 and AST23-2KO) were provided and generated by Dr. Tilo Kunath from The University of Edinburgh. The parental cell line AST23 was first described in Devine et al.19. The generation of the SNCA-corr cell line AST23-2KO is described in Chen et al.63. Both cell lines are deposited as EDi001-A and EDi001-A-5 in the hPSCreg repository, respectively (https://hpscreg.eu). The Miro1 R272Q and isogenic control cell lines originate from our laboratory and are described by Chemla et al.37. iPSCs were cultured on Geltrex-coated (Thermo Fisher Scientific) dishes in StemMACS iPS-Brew XF (Miltenyi Biotech). The medium was changed daily, and cells were passaged twice a week using 0.5 mM EDTA in PBS (Thermo Fisher Scientific). Mycoplasma testing was performed twice per month. mDA neurons were differentiated using a modified protocol based on Kriks et al.64,65,66. In summary, iPSCs were initially placed on Geltrex-coated six-well plates or T75 flasks at a density of 2 × 105 cells/cm2 in StemMACS iPS-Brew XF with 10 µM Y-27,632 (Hiss). The following day, the medium was changed to KnockOut DMEM medium with KnockOut serum replacement (both from Thermo Fisher Scientific), supplemented with 200 nM LDN19318 (Axon Medchem) and 10 µM SB431542 (Biozol) for dual SMAD-inhibition. On the second day, 100 ng/mL Shh C24II (Miltenyi Biotech), 2 µM Purmorphamine (Miltenyi Biotec), 100 ng/mL FGF8 (Peprotech), and 3 µM CHIR99021 (Miltenyi Biotec) were also added to the medium. After a period of 5 days, the medium was gradually transitioned to Neurobasal medium (Thermo Fisher Scientific), and SB431542 was removed from the medium. Starting from day 7, cells were cultivated only in the presence of LDN19318 and CHIR99021. On day 11, cells were transferred to Neurobasal/B27/L-glutamine medium with only CHIR99021. On day 13, cells were moved onto Geltrex-coated dishes in Neurobasal/B27/L-glutamine medium with supplements of 20 ng/mL BDNF, 20 ng/mL GDNF (both from Cell Guidance Sys.), 220 µM L-ascorbic-acid (Sigma-Aldrich), 10 µM DAPT (Axon Medchem), 1 ng/mL TGF-βIII (Peprotech), 0.5 mM dibutyryl-cAMP (Enzo Life Sciences), and 10 µM Y-27,632 (Hiss). Cells were kept in the same medium but without Y-27,632. Around days 23–25, cells were separated using StemPro Accutase (Thermo Fisher Scientific) and placed at a density of 1.4 × 105 cells/cm2 onto Geltrex-coated dishes. To remove non-neuronal cells, cultures were exposed to 1 µg/mL Mitomycin C for 2 h on day 26. On day 30, neuronal cultures were separated using StemPro Accutase with 10 µM Y-27,632 and singularized. Cells were counted and preserved at 2.5 × 106 cells/vial in CryoStor CS 10 tubes (Sigma-Aldrich).
Neuronal culture
30 days old mDA neurons were thawed and spun (400 g, 5 min, RT) in a basal medium (Table S2) supplemented with a ROCK inhibitor (Tocris #1254). The cell pellets were then reconstituted in a differentiation medium (Table S2) also supplemented with a ROCK inhibitor. 384-well plates (Perkin Elmer, Phenoplate poly-D-lysine-coated #6057500) were prepared with a coating of 10 µg/ml laminin at 4 °C overnight. Using trypan blue (Invitrogen, #T10282) and a Countess automated cell counter (Invitrogen), cells were seeded in the 384-well plates at a density of 10 × 103 cells/well. Edge wells were not used for seeding and were filled with PBS instead. Typically, the thawed cells were incubated at 37 °C and 5% CO2 for 6 days until they reached day 36, with changes in the differentiation medium occurring at D33. The processes of plate coating, cell seeding, and medium changes were automated using an Bravo Automated Liquid Handling Platform with 384ST head (Agilent) and a Multidrop Combi Dispenser (ThermoFischer Scientific).
Compound library and compound treatment
1,020 small molecules were selected from a bioactive compound library (Selleckchem), based on PD-relevant pathobiology, FDA approval status, and blood-brain-barrier penetrance (Table S3). Prostratin treatment: Neurons were cultured until D33, treated with 5 µM Prostratin (Biotechne) or DMSO for 72 h and immunofluorescently stained. Primary screening: Cryopreserved mDA D30 neurons were cultured in 384-well plates until D33, treated with each compound at a concentration of 5 µM for a duration of 72 h, and subsequently stained. Each compound underwent testing in four technical replicates. Compound wells were randomly assigned between technical replicates. 36 wells per plate were designated as control wells. The control conditions were used to establish the baseline and maximum effect profiles, specifically with SNCA-4x + DMSO serving as the baseline and SNCA-corr + DMSO as the maximum effect profile. Compounds and DMSO were dispensed using a Bravo Automated Liquid Handling Platform with 384ST head (Agilent). Dose-response testing: Identical to primary screening but with six different compound concentrations ranging from 0.25 to 20 µM. Biochemical assays: > D30 neurons were treated with 0.5–2.5 µM compound for 72 h. Microelectrode array (MEA): Ca. D60-old mDA neurons were treated up to 72 h with 2.5 µM Tyrphostin A9. Western blotting: > D50 mDA neurons were treated with 0.3 and 2.5 µM Tyrphostin A9 for 5 days.
Immunofluorescent staining
Neurons were fixed using 4% PFA (EMS Euromedex, #15710) for a duration of 20 min, followed by three PBS washes (Gibco, #14190136). The cells were then permeabilized and blocked using a solution of 10% FBS (Sigma, #F7524) and 0.1% Triton X-100 (Sigma, #T9284) in PBS for an hour. Primary antibodies against TH, αSyn, and MAP2 (Table S2) were prepared in an antibody dilution buffer (PBS with 5% FBS and 0.1% Triton X-100) and left to incubate with the cells at 4 °C overnight. This was followed by three PBS washes. Secondary antibodies and Hoechst (Table S2) were added to the cells in the antibody dilution buffer and left for an hour at room temperature, followed by another three PBS washes. Immunofluorescent staining was automated using an EL406 plate washer and dispenser (Biotek).
Image acquisition and feature extraction
Imaging was performed on a Yokogawa CV7000 microscope in scanning confocal mode using a dual Nipkow disk and a cooled sCMOS camera with 2,560 × 2,160 pixels and a pixel size of 6.5 μm without pixel binning. The system’s CellVoyager software (Yokogawa, version R1.17.05), the 405/488/561/640-nm solid laser lines, and a dry 40X objective were used to acquire 16 images in a 4 × 4 orientation from three 1 μm-separated Z-planes from the center of each well. Wells were visited in a row-wise “zigzag” fashion. Maximum intensity projections were performed. Image segmentation and feature extraction was performed with our in-house software PhenoLink (Ksilink, version 24.01.R0, https://github.com/Ksilink/PhenoLink). Image segmentation was performed on 4 × 4 stitched and illumination corrected images. Using intensity and size thresholds specific to each of the four imaging channels (Hoechst, TH, αSyn, and MAP2) we extracted a total of 124 morphological features, including count, texture, context, intensity and shape features (Table S4). We used location constraints to identify subcellular compartments. For example, by subtracting the segmented nucleus from the whole cell we obtained the cytoplasm or by calculating a 4 pixel-wide area around the outer edge of the cell we estimated the location of the cell membrane.
Data processing & morphological profile generation
All data manipulation and transformation processes were carried out using Python’s Pandas package67. All plots were generated using the Seaborn package68. All data processing and plotting can be reproduced using the provided Python Jupyter Notebooks and accompanying raw data. Briefly, raw morphological profiles were created per well by using the information from all measured 127 morphological features (Table S4). First, per feature a mean feature value from all acquired 16 image fields per well was calculated. The feature values were subjected to a standard scaling process using the StandardScaler function from the sklearn.preprocessing module in Python, resulting in the creation of 127-feature raw morphological profiles per well69. To determine the most defining morphological features per treatment condition, the morphological profiles were grouped based on their experimental replicates and treatments. The absolute differences in the mean values of the grouped data were calculated and followed by the computation of the mean and standard deviation of these absolute differences. We employed Principal Component Analysis (PCA) to compare morphological feature profiles across biological conditions and treatments. PCA was performed with 10 components and the explained variance ratio for each of the principal components was calculated. To measure how each original feature contributed to the principal components, the correlation between the original features and the first 10 principal components was calculated.
Primary screening data processing and hit selection
The raw data was composed out of 307,200 images from 4,800 wells distributed over 20 384-well plates. For all fluorescent channel images (Hoechst, TH, αSyn, and MAP2) we performed image segmentation and feature extraction, giving rise to a raw 127-feature morphological profile. Per plate, extreme outlier features were eliminated using a robust standard deviation-based method. This involved calculating the absolute difference between each data point and the feature median, and retaining data that was within 4σ of the Median Absolute Deviation (MAD). Next, data was normalized to allow comparisons between plates. We applied a robust Z-score normalization, during which the original data was adjusted based on its deviation from the median of the DMSO control data and was scaled by the MAD of the DMSO control data. This method is robust to outliers as it uses median and MAD, which are less sensitive to extreme values. Feature reduction was performed in two separate steps. Since the morphological profile contains an overrepresentation of texture features, first the two most distinguishing texture features were selected by calculating the z-factors for each feature, sorting them in descending order, and picking the top two. In the second step, features with less than 95% pairwise Pearson correlation were retained. In total, 40 features remained (Figure S2B). Plate quality was assessed using the Strictly Standardized Mean Difference (SSMD) based on the 40-feature Mahalanobis distance of SNCA-4x + Prostratin-treated control wells to SNCA-4x + DMSO control wells. SSMD values for all 20 plates were <-1, indicating a good plate quality. We then calculated the 40-feature Mahalanobis distances of all SNCA-4x compound-treated wells to the SNCA-4x + DMSO and SNCA-corr + DMSO control wells. We applied a Mahalanobis distance threshold of > 4 to identify compounds inducing morphological profiles different from SNCA-4x + DMSO control profiles. Additionally, we selected compounds based on close Mahalanobis distances to SNCA-corr + DMSO control profiles, thereby selecting compounds that induce morphological profiles similar to healthy control neurons. All data processing and hit selection steps can be reproduced by using the provided Python Jupyter Notebooks.
Evaluation of mitochondrial membrane potential
Mitochondrial membrane potential (Δψm) was estimated using the fluorescent probe JC-1 (ThermoFisher, T3168). Briefly, 60 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear-bottom 96-well cell culture plate and treated with selected compounds for three days. On neuronal maturation day 33, the neuronal medium was removed and cells were incubated with 100µL/well of fresh neuronal medium containing 10 µΜ of JC-1 dye for 30 min at 37 °C. After incubation, cells were washed three times with pre-warmed DPBS (ThermoFisher, #14190094). Fluorescence was measured using a Cytation 5 (Bio-Tek) fluorescence plate reader, with excitation at 560 nm and emission at 610 nm for the aggregated JC-1 and excitation at 500 nm and emission at 550 nm for the monomeric JC-1. Mitochondrial membrane potential was estimated by calculating the ratio of aggregated to monomeric JC-1 fluorescent signal. Statistical analyses were conducted on data from, at least, three independent biological replicates.
Measurement of intracellular ROS and glutathione levels
Intracellular ROS levels were measured using the ROS-ID Total ROS/Superoxide detection kit (Enzo, Enz-51010), following the manufacturer’s instructions. Briefly, 60 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear bottom 96-well cell culture plate. Neurons were treated for three days with the selected compounds. On neuronal maturation day 33, the neuronal medium was removed and the cells were washed once with pre-warmed DPBS (ThermoFisher, #14190094). Cells were then incubated with 100 µL/well of ROS Detection Solution for 1 h at 37 °C. Finally, fluorescence is measured without removing the detection solution, using a Cytation 5 (Bio-Tek) fluorescence plate reader (bottom reading at Ex = 490 nm, Em = 520 nm). Fluorescence intensity was normalized to total cell protein levels, using the BCA Protein Detection Kit (ThermoFisher, #23227). Statistical analyses were conducted on data from three independent biological replicates. Simultaneously, intracellular GSH levels were measured using the monochlorobimane (MCB, AAT Bioquest, #76421-73-3) assay. In this procedure, 60 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear-bottom 96-well cell culture plate and treated with selected compounds for three days. On neuronal maturation day 33, the neuronal medium was removed, and the cells were washed once with pre-warmed DPBS (ThermoFisher, #14190094). The cells were then incubated with 100 µL per well of a detection solution containing 50 µM MCB for 1 h a, with excitation at 360 nm and emission at 480 nm t 37 °C. After incubation, fluorescence was measured directly without removing the detection solution using a Cytation 5 (Bio-Tek) fluorescence plate reader, with excitation at 360 nm and emission at 480 nm. The fluorescence intensity readings were normalized to total cell protein levels, which were determined using the BCA Protein Detection Kit (ThermoFisher, #23227). Statistical analyses were conducted on data from three independent biological replicates.
Evaluation of cytotoxicity
Cytotoxicity was estimated using the lactate dehydrogenase (LDH) assay (Canvax, CA0020), following the manufacturer’s instructions. 60 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear bottom 96-well cell culture plate. Neurons were treated for three days with the selected compounds. On neuronal maturation day 33, 50 µL of media was obtained from each well and transferred to a 96-well cell culture plate, in order to evaluate the release of LDH into the cell culture media. It was added to an equal volume (50 µL) of mixed LSH assay buffer and dye solution and was incubated for 30 min, at room temperature, protected from light. Finally, 50 µL Stop Solution was added per well and the absorbance (OD) was measured using a Cytation 5 (Bio-Tek) plate reader (450 nm). LDH absorbance was normalized to total cell protein levels, using the BCA Protein Detection Kit (ThermoFisher, #23227). Statistical analyses were conducted on data from three independent biological replicates.
Measurement of intracellular ATP levels
Intracellular ATP levels were evaluated using the CellTiter-Glo 2.0 (Promega, G9241), according to the product’s instructions. Briefly, 70 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear-bottom 96-well cell culture plate and treated with selected drugs or compounds for three days. On neuronal maturation day 33, the neuronal medium was removed, and cells were washed once pre-warmed DPBS (ThermoFisher, #14190094). Cells were incubated with 75 µL per well of HBSS and an equal volume of the CellTiter-Glo 2.0 mix was added in each well. The plates were shaken for 2 min in RT, followed by 10 min of incubation in RT. Luminescence was measured using a Cytation 5 (Bio-Tek) plate reader. Luminescence intensity was normalized to total cell protein levels, using the BCA Protein Detection Kit (ThermoFisher, #23227). Statistical analyses were conducted on data from three independent biological replicates.
Analysis of mitochondrial respiration
Oxygen consumption rates (OCRs) were measured using the Seahorse XFe96 FluxPak (Agilent, #102416-100) and the XF Cell Mito Stress Test Kit (Agilent, #103015-100), following the manufacturer’s instructions. For this experiment, 70 × 103 mDA neurons were seeded per well in a Geltrex-coated 96-well Seahorse cell culture plate. On neuronal maturation day 33, the neuronal medium in each well was replaced with 175 µl of Seahorse XF DMEM Medium (pH 7.4, Agilent, #103575-100) containing 25 mM Glucose, 2 mM L-Glutamine, and 1 mM Sodium Pyruvate. The cells were incubated for 1 h at 37 °C without CO2. Meanwhile, the sensor cartridge, pre-equilibrated overnight in XF Calibrant solution at 37 °C without CO2, was prepared with standard mitochondrial inhibitors: Oligomycin (2.5 µM), FCCP or tyrphostin analogs (2.5 µM), and a mixture of Rotenone and Antimycin A (1 µM each). After a further 30-minute incubation at 37 °C without CO2, the sensor cartridge and the cell culture plate were loaded into the XFe96 Extracellular Flux Analyzer (Agilent). OCR was measured three times for basal respiration and after each injection of mitochondrial toxins. Post-assay, the cell culture medium was removed, and total protein content was determined using a BCA Protein Assay Kit (ThermoFisher, #23227). The OCR data were analyzed with XFe Wave Pro software (Agilent, version 10_3_0_515), normalized to total cell protein levels (based on the BCA assay), and exported using the Seahorse XF Cell Mito Stress Test Report Generator. Statistical analyses were conducted on data from three independent biological replicates. The study calculated various parameters related to oxidative phosphorylation, including basal respiration, proton leakage across the inner mitochondrial membrane, ATP-coupled respiration, maximal respiration, and spare respiratory capacity.
Multi-electrode array (MEA)
Cryopreserved 30 days old neurons were thawed in a water bath and centrifuged (400 g, 5 min, RT) in basal medium (Table S2) supplemented with ROCK inhibitor (Tocris, #1254). Cell pellets were resuspended in differentiation medium (Table S2) supplemented with ROCK inhibitor. 48-well CytoView MEA plates (Axion Biosystems, #M768-tMEA-48 W) were coated with 15 µg/ml Polyethyleneimine solution (Sigma, #P3143) for 1 h at 37 °C followed by 10 µg/ml Laminin coating for 2 h at 37 °C. Using Tryphan Blue (Sigma, # T8154-20ML) and a Countess automated cell counter (Invitrogen), 80 × 103 cells/well were seeded in the CytoView plate. Cells were incubated at 37 °C and 5% CO2 for 26–33 days with differentiation medium changes twice a week. Once a week, 1 µg/ml Laminin was added to the fresh media to maintain cell attachment. Neurons were then treated with compound for 72 h. The electrical activity was recorded on the Maestro Pro multiwell MEA (Axion Biosystems) using the AxIS Navigator software (version 3.5.1, Axion Biosystems). Electrical activity was recorded for 10 min at a 12.5 kHz sampling rate and a 5.5 standard deviation threshold level for action potential detection. Before plate loading the device was allowed to equilibrate for approximately 30 min to 37 °C and 5% CO2. Data analyses were performed using the Neural Metric Tool (version 3.1.7, Axion Biosystems). Activity is expressed as the mean firing rate across electrodes in a well.
SDS gel electrophoresis and Western blotting
1.5 × 106 mDA neurons were cultured for at least 50 days in Geltrex-coated 6-well plates with neuronal media and treated with compound for five days. At day 50, the neuronal media was removed and cells were washed with pre-warmed DPBS. For high resolution αSyn detection, the following steps were performed: DPBS was removed and cells were crosslinked by adding 1 mL pre-warmed PBSx1 supplemented with 2 mM disuccinimidyl glutarate (DSG, ThermoFisher, #20593) and protease inhibitors for 30 min at 37 °C. Crosslinking was terminated by adding 50 mM Tris-HCl, pH = 7.5 for 15 min at room temperature. Cells were collected and centrifuged for 10 min at 1000xg at room temperature. The supernatant was removed and the cell pellet was lysed using lysis buffer (100 mM NaCl, 5 mM EDTA, 50 mM Tris-Cl, pH 7.4, 50 mM NaF, containing 1% (v/v) Igepal CA-630 and supplemented with protease inhibitors). Cells were incubated for 30 min at 4 °C with the lysis buffer and sonicated three times. Total cellular protein was measured using the BCA Protein Detection Kit (ThermoFisher, #23227) and 25 µg of cell lysate were electrophoresed at 100 V constant on NuPAGE 4–12% Bis-Tris Midi gels (Invitrogen, NP0321BOX) with NuPage MES-SDS running buffer (Invitrogen, NP0002). After electrophoreses, gels were electroblotted in nitrocellulose membranes, using a wet-transfer approach. After electroblotting, the membranes were briefly rinsed in ultrapure water and incubated in 4% paraformaldehyde/PBS for 30 min at room temperature, followed by blocking for 1 h at room temperature in TBS-Tween 0.1% (v/v) supplemented with 5% (w/v) dry milk. To visualize other proteins and αSyn at lower size resolution, the following steps were performed: Cellular pellets were lysed using RIPA buffer, including phosphatase and protease inhibitors, and total protein was quantified using the Pierce BCA assay kit, according to the manufacturer’s instructions. 30 mg of total exosomal extract was subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis on 8–12% Bis-Tris gel (Thermo Fisher Scientific). The separated proteins were transferred onto a nitrocellulose membrane, followed by blocking with 5% non-fat milk powder (w/v) in Tris-buffered saline (10 mM Tris-HCl, pH 7.5, 100mM NaCl, 0.1% Tween 20) for 1 h at room temperature. After blocking, membranes were incubated with the primary antibodies listed in Table S2. The next day, membranes were washed 3 × 10 min in PBS-Tween 0.1% and incubated with the corresponding secondary antibodies (Table S2) at RT for 1 h in the dark. Membranes were rinsed in PBS-Tween 0.1%, washed at least 3 × 10 min in PBS-Tween 0.1%, incubated for 3 min with ECL western blotting substrate, and imaged on a LI-COR Odyssey XF imaging system.
Measurement of intracellular calcium levels
Intracellular calcium levels were evaluated using the cell-permeant calcium indicator Rhod-FF AM (Invitrogen, R23983). Briefly, 70 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear-bottom 96-well cell culture plate and treated with the selected compound for three days. On neuronal maturation day 33, the neuronal medium was removed, and cells were washed with the pre-warmed HBS buffer (145 mM NaCl, 5 mM KCl, 1 mM CaCl2, 1 mM MgCl2, 10 mM HEPES, 10mM glucose) and incubated for 1 h at room temperature, protected from light, with HBS buffer containing 0.1% (w/v) BSA and 5 µΜ Rhos-FF AM. After incubation, the cells were washed twice with 100µL per well HBS and were incubated for 20 min at room temperature, protected from light, for dye de-esterification. Fluorescence was measured 4 times, every 30 s, using a Cytation 5 (Bio-Tek) fluorescence plate reader, with excitation at 550 nm and emission at 580 nm, at 28 °C. After the first measurement, HBS buffer supplemented with 10 µΜ of Ionomycin (Sigma, #19657) was added to the cells, and the plate was measured again (8 times (8 times, every 30 s). Fluorescence intensity was normalized to total cell protein levels, using the BCA Protein Detection Kit (ThermoFisher, #23227).
Evaluation of proteasomal activity
The proteasomal activity was evaluated by measuring the chymotrypsin-like and caspase-like activity. The chymotrypsin-like activity was measured using the fluorogenic substrate Suc-Leu-Leu-Val-Tyr-AMC (Enzo, BML-PB02-0005) and the caspase-like activity was measured by using the fluorogenic substrate Z-Leu-Leu-Glu-AMC (Enzo, BML-ZW9345-0005). Briefly, 60 × 103 mDA neurons were seeded per well in a Geltrex-coated black, clear-bottom 96-well cell culture plate and treated with the selected compound for three days. On neuronal maturation day 33, the neuronal medium was removed and cells were washed with the pre-warmed assay buffer (20 mM Tris-HCl pH = 7.5, 5 mM MgCl2, 0.1 mM EDTA, 1 mM DTT, 1 mM ATP, pH = 7.4). After washing, cells are incubated for 30 min at 37 °C with 100 µL per well of assay buffer containing 100 µΜ of the respective fluorescent substrate. Finally, fluorescence was measured using a Cytation 5 (Bio-Tek) fluorescence plate reader, with excitation at 380 nm and emission at 460 nm, at 37 °C. Fluorescence intensity was normalized to total cell protein levels, using the BCA Protein Detection Kit (ThermoFisher, #23227). Statistical analyses were conducted on data from three independent biological replicates.
qPCR
Cellular pellets were used for RNA isolation using the RNeasy Plus Mini kit (Qiagen), following the manufacturer’s instructions. The isolated RNA was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), following the manufacturer’s instructions. Gene expression was assessed on a LightCycler 480 II (Roche Diagnostics), using the PowerTrack SYBR Green Mastermix (Thermo Fisher Scientific), following the manufacturer’s instructions. The sets of primers that were used are shown in TableS2. SNCA fold changes were quantified relative to ACTB using the ∆∆Ct approach.
Structural analogues
Compounds were directly obtained from Selleckchem (CIDs: 5614, 2044, 1838043, 135419190), Enamine (CID: 135407924), or custom synthesized by SynHet (CID: 3511020). For compound CID 3511020, we chose the abbreviation CEPPE based on its IUPAC name 2-cyano-3-(3-ethoxy-4-propan-2-yloxyphenyl)prop-2-enethioamide. Chemical structures were plotted from their Simplified molecular-input line-entry system (SMILES) strings using Python. The rdkit package was used to create molecule objects from the SMILES strings and to compute 2D coordinates for these molecules. Tanimoto coefficients were calculated using their CID code via the PubChem Score Matrix service available at https://pubchem.ncbi.nlm.nih.gov/score_matrix/score_matrix.cgi.
Data availability
The datasets supporting the conclusions of this article are available in the GitHub repository, https://github.com/Ksilink/Notebooks/tree/main/Neuro/PD_MorphProfileScreening. Raw data is available as CSV files which can be processed using the provided Python-based Jupyter Notebooks.
References
Wüllner, U. et al. The heterogeneity of Parkinson’s disease. J. Neural Transm. 130, 827–838 (2023).
Poewe, W. et al. Parkinson disease. Nat. Rev. Dis. Primer. 3, 17013 (2017).
Coukos, R. & Krainc, D. Key genes and convergent pathogenic mechanisms in Parkinson disease. Nat. Rev. Neurosci. 25, 393–413 (2024).
Espay, A. J. et al. Biomarker-driven phenotyping in parkinson’s disease: A translational missing link in disease‐modifying clinical trials. Mov. Disord. 32, 319–324 (2017).
Krüger, R. et al. Classification of advanced stages of parkinson’s disease: translation into stratified treatments. J. Neural Transm. 124, 1015–1027 (2017).
Prasuhn, J. & Brüggemann, N. Genotype-driven therapeutic developments in parkinson’s disease. Mol. Med. 27, 42 (2021).
Laperle, A. H. et al. iPSC modeling of young-onset parkinson’s disease reveals a molecular signature of disease and novel therapeutic candidates. Nat. Med. 26, 289–299 (2020).
Vuidel, A. et al. High-content phenotyping of parkinson’s disease patient stem cell-derived midbrain dopaminergic neurons using machine learning classification. Stem Cell. Rep. 17, 2349–2364 (2022).
D’Sa, K. et al. Prediction of mechanistic subtypes of parkinson’s using patient-derived stem cell models. Nat. Mach. Intell. 5, 933–946 (2023).
Bose, A., Petsko, G. A. & Studer, L. Induced pluripotent stem cells: a tool for modeling parkinson’s disease. Trends Neurosci. 45, 608–620 (2022).
Schiff, L. et al. Integrating deep learning and unbiased automated high-content screening to identify complex disease signatures in human fibroblasts. Nat. Commun. 13, 1590 (2022).
Cholerton, B. et al. Precision medicine. Am. J. Pathol. 186, 500–506 (2016).
Espay, A. J., Brundin, P. & Lang, A. E. Precision medicine for disease modification in Parkinson disease. Nat. Rev. Neurol. 13, 119–126 (2017).
Haston, K. M. & Finkbeiner, S. Clinical trials in a dish: the potential of pluripotent stem cells to develop therapies for neurodegenerative diseases. Annu. Rev. Pharmacol. Toxicol. 56, 489–510 (2016).
Polymeropoulos, M. H. et al. Mutation in the α-Synuclein gene identified in families with parkinson’s disease. Science 276, 2045–2047 (1997).
Krüger, R. et al. AlaSOPro mutation in the gene encoding α-synuclein in parkinson’s disease. Nat. Genet. 18, 106–108 (1998).
Singleton, A. B. et al. α-Synuclein locus triplication causes parkinson’s disease. Science 302, 841–841 (2003).
Chartier-Harlin, M. C. et al. α-synuclein locus duplication as a cause of Familial parkinson’s disease. Lancet 364, 1167–1169 (2004).
Devine, M. J. et al. Parkinson’s disease induced pluripotent stem cells with triplication of the α-synuclein locus. Nat. Commun. 2, 440 (2011).
Byers, B. et al. SNCA triplication parkinson’s patient’s iPSC-derived DA neurons accumulate α-Synuclein and are susceptible to oxidative stress. PLoS ONE. 6, e26159 (2011).
Mohamed, N. V. et al. Midbrain organoids with an SNCA gene triplication model key features of synucleinopathy. Brain Commun. 3, fcab223 (2021).
Iannielli, A. et al. Modeling native and seeded synuclein aggregation and related cellular dysfunctions in dopaminergic neurons derived by a new set of isogenic iPSC lines with SNCA multiplications. Cell. Death Dis. 13, 881 (2022).
Oliveira, L. M. A. et al. Elevated α-synuclein caused by SNCA gene triplication impairs neuronal differentiation and maturation in parkinson’s patient-derived induced pluripotent stem cells. Cell. Death Dis. 6, e1994–e1994 (2015).
Zambon, F. et al. Cellular α-synuclein pathology is associated with bioenergetic dysfunction in parkinson’s iPSC-derived dopamine neurons. Hum. Mol. Genet. 28, 2001–2013 (2019).
Heman-Ackah, S. M. et al. Alpha-synuclein induces the unfolded protein response in parkinson’s disease SNCA triplication iPSC-derived neurons. Hum. Mol. Genet. 26, 4441–4450 (2017).
Burbulla, L. F. et al. Dopamine oxidation mediates mitochondrial and lysosomal dysfunction in parkinson’s disease. Science 357, 1255–1261 (2017).
Angelova, P. R. et al. Alpha synuclein aggregation drives ferroptosis: an interplay of iron, calcium and lipid peroxidation. Cell. Death Differ. 27, 2781–2796 (2020).
Tanudjojo, B. et al. Phenotypic manifestation of α-synuclein strains derived from parkinson’s disease and multiple system atrophy in human dopaminergic neurons. Nat. Commun. 12, 3817 (2021).
Chandrasekaran, S. N., Ceulemans, H., Boyd, J. D. & Carpenter, A. E. Image-based profiling for drug discovery: due for a machine-learning upgrade? Nat. Rev. Drug Discov. 20, 145–159 (2021).
Chandrasekaran, S. N. et al. JUMP Cell Painting Dataset: Morphological Impact of 136,000 Chemical and Genetic Perturbations. http://biorxiv.org/lookup/doi/10.1101/2023.03.23.534023. 10.1101/2023.03.23.534023 (2023).
Bray, M. A. et al. Cell painting, a high-content image-based assay for morphological profiling using multiplexed fluorescent dyes. Nat. Protoc. 11, 1757–1774 (2016).
Park, S. J. et al. A receptor tyrosine kinase inhibitor, tyrphostin A9 induces cancer cell death through Drp1 dependent mitochondria fragmentation. Biochem. Biophys. Res. Commun. 408, 465–470 (2011).
Sagara, Y., Ishige, K., Tsai, C. & Maher, P. Tyrphostins protect neuronal cells from oxidative stress. J. Biol. Chem. 277, 36204–36215 (2002).
Ferreira, A. F. F. et al. Inhibition of TRPM2 by AG490 is neuroprotective in a parkinson’s disease animal model. Mol. Neurobiol. 59, 1543–1559 (2022).
Dai, X. et al. Tyrphostin A9 protects axons in experimental autoimmune encephalomyelitis through activation of ERKs. Life Sci. 294, 120383 (2022).
Iaubasarova, I. R. et al. The mitochondria-targeted derivative of the classical uncoupler of oxidative phosphorylation carbonyl cyanide m-chlorophenylhydrazone is an effective mitochondrial recoupler. PLOS ONE. 15, e0244499 (2020).
Chemla, A. et al. Generation of two induced pluripotent stem cell lines and the corresponding isogenic controls from parkinson’s disease patients carrying the heterozygous mutations c.815G > A (p.R272Q) or c.1348C > T (p.R450C) in the RHOT1 gene encoding Miro1. Stem Cell. Res. 71, 103145 (2023).
Mizushima, N. & Yoshimori, T. How to interpret LC3 Immunoblotting. Autophagy 3, 542–545 (2007).
McDiarmid, A. H. et al. Morphological profiling in human neural progenitor cells classifies hits in a pilot drug screen for Alzheimer’s disease. Brain Commun. 6, fcae101 (2024).
Han, X. et al. Whole human-brain mapping of single cortical neurons for profiling morphological diversity and stereotypy. Sci. Adv. 9, eadf3771 (2023).
Peng, H. et al. Morphological diversity of single neurons in molecularly defined cell types. Nature 598, 174–181 (2021).
Kilfeather, P. et al. Single-cell Spatial transcriptomic and Translatomic profiling of dopaminergic neurons in health, aging, and disease. Cell. Rep. 43, 113784 (2024).
Kim, G. H. J. et al. A zebrafish screen reveals Renin-angiotensin system inhibitors as neuroprotective via mitochondrial restoration in dopamine neurons. eLife 10, e69795 (2021).
Cui, J., Carey, J. & Reijo Pera, R. A. Identification of DOT1L inhibitor in a screen for factors that promote dopaminergic neuron survival. Front. Aging Neurosci. 14, 1026468 (2022).
Tabata, Y. et al. T-type calcium channels determine the vulnerability of dopaminergic neurons to mitochondrial stress in Familial Parkinson disease. Stem Cell. Rep. 11, 1171–1184 (2018).
Yamaguchi, A. et al. Identifying therapeutic agents for amelioration of mitochondrial clearance disorder in neurons of Familial Parkinson disease. Stem Cell. Rep. 14, 1060–1075 (2020).
Asantewaa, G. et al. Glutathione synthesis in the mouse liver supports lipid abundance through NRF2 repression. Nat. Commun. 15, 6152 (2024).
Cvetko, F. et al. Nrf2 is activated by disruption of mitochondrial thiol homeostasis but not by enhanced mitochondrial superoxide production. J. Biol. Chem. 296, 100169 (2021).
Harvey, C. J. et al. Nrf2-regulated glutathione recycling independent of biosynthesis is critical for cell survival during oxidative stress. Free Radic Biol. Med. 46, 443–453 (2009).
von Willebrand, M., Zacksenhaus, E., Cheng, E., Glazer, P. & Halaban, R. The tyrphostin AG1024 accelerates the degradation of phosphorylated forms of retinoblastoma protein (pRb) and restores pRb tumor suppressive function in melanoma cells. Cancer Res. 63, 1420–1429 (2003).
Shvets, E., Fass, E., Scherz-Shouval, R. & Elazar, Z. The N-terminus and Phe52 residue of LC3 recruit p62/SQSTM1 into autophagosomes. J. Cell. Sci. 121, 2685–2695 (2008).
Shvets, E., Abada, A., Weidberg, H. & Elazar, Z. Dissecting the involvement of LC3B and GATE-16 in p62 recruitment into autophagosomes. Autophagy 7, 683–688 (2011).
Bjørkøy, G., Lamark, T. & Johansen, T. p62/SQSTM1: A missing link between protein aggregates and the autophagy machinery. Autophagy 2, 138–139 (2006).
Walrafen, P. et al. Both proteasomes and lysosomes degrade the activated erythropoietin receptor. Blood 105, 600–608 (2005).
Terada, H. & Van Dam, K. On the, stoichiometry between uncouplers of oxidative phosphorylation and respiratory chains. The catalytic action of SF 6847 (3,5-di-tert-butyl-4-hydroxybenzylidenemalononitrile). Biochim. Biophys. Acta BBA - Bioenerg. 387, 507–518 (1975).
Geisler, J. G., Marosi, K., Halpern, J. & Mattson, M. P. DNP, mitochondrial uncoupling, and neuroprotection: A little dab’ll do Ya. Alzheimers Dement. 13, 582–591 (2017).
Zorov, D. B. et al. Neuroprotective potential of mild uncoupling in mitochondria. Pros and cons. Brain Sci. 11, 1050 (2021).
Terada, H., Fukui, Y., Shinohara, Y. & Ju-ichi, M. Unique action of a modified weakly acidic uncoupler without an acidic group, methylated SF 6847, as an inhibitor of oxidative phosphorylation with no uncoupling activity: possible identity of uncoupler binding protein. Biochim. Biophys. Acta BBA - Bioenerg. 933, 193–199 (1988).
Peruzzotti-Jametti, L. et al. Mitochondrial complex I activity in microglia sustains neuroinflammation. Nature 628, 195–203 (2024).
Gao, C., Jiang, J., Tan, Y. & Chen, S. Microglia in neurodegenerative diseases: mechanism and potential therapeutic targets. Signal. Transduct. Target. Ther. 8, 359 (2023).
Park, D. S. et al. iPS-cell-derived microglia promote brain organoid maturation via cholesterol transfer. Nature 623, 397–405 (2023).
Zhang, W. et al. Microglia-containing human brain organoids for the study of brain development and pathology. Mol. Psychiatry. 28, 96–107 (2023).
Chen, Y. et al. Engineering synucleinopathy-resistant human dopaminergic neurons by CRISPR ‐mediated deletion of the SNCA gene. Eur. J. Neurosci. 49, 510–524 (2019).
Kriks, S. et al. Dopamine neurons derived from human ES cells efficiently engraft in animal models of parkinson’s disease. Nature 480, 547–551 (2011).
Ryan, S. D. et al. Isogenic human iPSC parkinson’s model shows nitrosative Stress-Induced dysfunction in MEF2-PGC1α transcription. Cell 155, 1351–1364 (2013).
Weykopf, B. et al. Induced pluripotent stem cell-based modeling of mutant LRRK2-associated parkinson’s disease. Eur. J. Neurosci. 49, 561–589 (2019).
McKinney, W. Data structures for statistical computing in python. Proc. 9th Python Sci. Conf. 56-61 https://doi.org/10.25080/Majora-92bf1922-00a (2010).
Waskom, M. L. Seaborn: statistical data visualization. J. Open. Source Softw. 6, 3021 (2021).
Pedregosa, F. et al. Scikit-learn: machine learning in python. J. Mach. Learn. Res. 12, 2825–2830 (2011).
Acknowledgements
We thank Dr. Tilo Kunath (University of Edinburgh) for providing the AST23 and AST23-2KO iPSC lines. We are grateful for the constant discussions with various colleagues at Ksilink and the Luxembourg Centre for Systems Biomedicine (LCSB) which improved this work.
Author information
Authors and Affiliations
Contributions
V.G. designed biochemical experiments, performed neuronal differentiation, biochemical assays, and data analysis; A.We. designed imaging experiments, performed neuronal culturing, imaging assay automation, and compound screening; L.C. developed and performed image analysis and data processing; D.H. performed MEA experiments; K.S. analyzed the MEA data; A.O. co-supervised data processing, P.A.B. co-initiated the study and contributed to grant writing; B.F.R.S. contributed to biochemical experiments; I.B. co-designed biochemical experiments; A.Wi., A.Z. and O.P. identified and provided Tyrphostin A9 structural analogues; R.K. co-initiated the study and supervised biochemical experiments; P.S. co-initiated the study and supported assay design; J.H.W. supervised the study, designed experiments, conceptualized and performed image analysis and data processing, and wrote the manuscript with input from all authors.
Corresponding author
Ethics declarations
Competing interests
A.We., L.C., D.H., K.S., A.O., P.S., and J.H.W. are or were employed by Ksilink. R.K. is receiving or has received research grants from the European Institute of Innovation and Technology (EIT Health), the Innovative Medicines Initiative (IMI) of the European Union and the European pharmaceutical industry, as well as personal speaker’s honoraria and/or travel grants from Abbvie, Zambon and Medtronic; R.K. participated as PI or site-PI for industry sponsored clinical trials without receiving honoraria. All other authors declare no conflict of interest.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Gorgogietas, V., Weiss, A., Cousin, L. et al. Morphological profiling reveals neuroprotection via mitochondrial uncoupling in human dopaminergic neurons. Sci Rep 15, 29507 (2025). https://doi.org/10.1038/s41598-025-14735-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-025-14735-0
