HEK293T (RRID:CVCL_0063) cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum. By analysis of 21 genetic loci the cells were validated in November 2017 (Promega, Power Plex 21 PCR Kit carried out by Eurofins MedigenomixForensiker, Ebersberg, Germany). Cells were plated on poly-L-Lysine coated glass coverslips or on plastic plates. Twenty-four hours after plating, cells were transfected using metafectene (Cat# T020-5.0 Biontex Laboratories, Munich, Germany) following the manufactures instructions.

Neurons derived from human-induced pluripotent stem cells (hiPSCs) were cultured and differentiated as described in Reinhardt et al. (2013), starting from a pre-differentiated neuronal progenitor state kindly provided by Jared Sterneckert. In short, progenitors were cultured in N2B27 medium (50:50 DMEM-F12 (Invitrogen, Paisley, UK)/Neurobasal (Invitrogen, Paisley, UK) with 1:200 N2 supplement (Invitrogen, Grand Island, Kentucky, USA), 1:100 B27 supplement lacking vitamin A (Invitrogen, Grand Island, Kentucky, USA) and 1% penicillin/streptomycin/glutamine (Life Technologies, Grand Island, Kentucky, USA). During expansion, N2B27 was freshly supplemented with 3 μM CHIR99021 (Cayman, Ann Arbor, Michigan, USA), 150 μM ascorbic acid (AA) (Sigma-Aldrich, St. Louis, Missouri, USA), and 0.5 μM smoothened agonist (SAG) (Cayman, Ann Arbor, Michigan, USA). For differentiation, 0.8*10⁶ cells were seeded on Matrigel-coated (Corning, New York, USA) 6-well plates (Corning, Kennebunk, Maine, USA) and cultured for 6 days in N2B27 supplemented with 1 ng/ml Brain-derived neurotropic factor (BDNF; PeproTech, Hamburg, Germany), 0.2 mM AA, 1 μM retinoic acid (Sigma-Aldrich, St. Louis, Missouri, USA), 1 ng/ml Glial-derived neurotrophic factor (GDNF; Sigma-Aldrich, St. Louis, Missouri, USA) and 0.5 μM SAG. For final maturation, cells were maintained in N2B27 with 0.1 mM dibutyrylcyclic adenosine monophosphate (Sigma-Aldrich, St. Louis, Missouri, USA), 2 ng/ml BDNF, 0.2 mM AA, 1 ng/ml TGFβ-3 (PeproTech, Hamburg, Germany) and 2 ng/ml GDNF. During the first 2 days of maturation, 5 ng/ml Activin A (Biomol, Hamburg, Germany) was added to the medium. On maturation day 2, cells were split in the final format: 50,000 cells/well, 96-well plates (Greiner bio-one, Kremsmünster, Austria) coated with poly-L-ornithine (Sigma-Aldrich, St. Louis, Missouri, USA) and 1% Laminin (Roche, Mannheim, Germany). One day after the final plating, neurons were transduced with lentiviral vector containing Synuclein-T2A-GFP. Twenty-one days after lentiviral transduction, cells were treated with ML-SA1 or DMSO.

LUHMES cells were cultured and differentiated as described before (Scholz et al., 2011; Neuhof et al., 2021). In short, Cells were cultured in culture dishes (Nunclon Delta EasY, Thermofisher, Waltham, Massachusetts, USA), precoated with 50 μg/ml Poly-L-ornithine (Sigma-Aldrich, St. Louis, Missouri, USA) and 1 μg/ml human plasma fibronectin (Gibco, Thermo Fisher Scientific, Waltham, USA) in distilled water at 37°C for 3 h. Undifferentiated LUHMES cells were cultured in Dulbecco’s modified Eagle’s medium/F12 (DMEM/F12, Gibco, USA), supplemented with 1× N-2 supplement (Gibco/Invitrogen, Thermo Fisher Scientific, Waltham, USA), L-Glutamin (Gibco/Invitrogen, USA) and 40 ng/ml human recombinant basic FGF (bFGF, Sigma-Aldrich, USA), at 37°C in 5% CO₂ incubator.

Cell differentiation was done as described before (Scholz et al., 2011; Neuhof et al., 2021). Briefly, 0.6× 10⁶ LUHMES cells were seeded onto a pre-coated T25 flask. 24 h later, medium change was performed with a differentiation medium containing DMEM/F12, N-2 supplement, L-Glutamine, 1 μg/ml tetracycline (Sigma-Aldrich, USA), 40 ng/ml human recombinant GDNF (PeproTech Germany, Hamburg, Germany) and 1 mmol/L dibutyryl cAMP (Sigma-Aldrich, USA). After 2 days of differentiation, cells were re-plated to pre-coated 96-well plates or 24-well plates and lentiviral transduction was performed 3 h after seeding. Medium change was performed after 24 h following every other day. On day 9 of differentiation, cells were treated with ML-SA1, ML-SI3, BafA1, or DMSO for 2 h or 24 h. Validation of dopaminergic differentiation was performed via immunostaining on day 0 and day 9 of differentiation (Supplementary Figures 4A,B).

We used 3rd-generation lentivirus as described before (Szego et al., 2019) containing full-length human α-synuclein cDNA on vector CD526A-1 followed by a T2A-GFP sequence under the vector promotor EF1α. For virus production a second-generation lentivirus packaging system was used, based on the plasmids pcziVSV-G and pCD/NL-BH, kindly provided by Prof. Dr. Dirk Lindemann. For the production of viral particles, HEK293T cells were plated in 10 cm dishes (Greiner bio-one, Kremsmünster, Austria) and transfected with both packaging plasmids and the corresponding transfer vector using PEI (Sigma-Aldrich, St. Louis, Missouri, USA). Viral supernatant was collected 48 h after transfection and concentrated by ultracentrifugation. The resulting viral particles were used to transduce mature neurons. To adjust neuronal infection efficiencies, viral particle infectious units were calculated by fluorescence microscopy of infected HEK293T cells. Experiments with lentivirus were performed to appropriate safety regulations in TU Dresden with the permit Az. 45-8452/120.

Unless noted otherwise, HEK293T cells were chemically treated 6 h after transfection for 18 h and subsequently processed and analyzed.

ML-SA1 (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) was dissolved in DMSO to a 25 mM stock solution. It was used in a final concentration of 25 μM for HEK cells and depicted concentrations for neurons. ML-SI3 (MedKoo Biosciences, Inc., Morrisville, USA) was dissolved in DMSO to a 25 mM stock solution using a final concentration on neurons as depicted. Inhibition of fusion of autophagosomes and lysosomes was achieved by incubation with 5 nM bafilomycin A1 (BafA1; Tocris Bioscience, Bristol, GB, stock solution of 1 μM in DMSO). For time-lapse imaging and Western blot analysis, 100 nM bafilomycin A1 was used. Blocking the formation of autophagosomes was achieved by incubation with 2 mM 3-methlyadenine (3-MA; Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany, stock solution 2M in DMSO). The selective Ca²⁺ chelator BAPTA-AM was used in concentrations of 0.1 μM up to 100 μM (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany, stock solution of 100 mM in DMSO). For chemically induced dimerization, 100 nM rapamycin (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) was used for 1 h before processing the cells. Activation of autophagy by rapamycin was achieved by incubation with 100 nM rapamycin for 18 h, 6 h after transfection. The PIKfyve-Inhibitor YM201636 was used in a concentration of 800 nM (Cayman Chemical Company, Michigan, US, stock solution 80 μM in DMSO). For all drugs dissolved in DMSO, an equal concentration of DMSO (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) was used as vehicle control.

A53T-α-synuclein tagged with EGFP by the interaction of a PDZ-binding motif with its PDZ domain was described previously (Opazo et al., 2008). A tandem-fluorescence version (mRFP-GFP) was used as described (Dinter et al., 2016). For the discrimination of cytosolic aggregates from vesicles by chemically-induced dimerization, we used A53T-α-synuclein fused to FRB (FKBP-rapamycin binding domain) and mCherry and in addition FK-506-binding protein (FKBP) fused to GFP (Dinter et al., 2016). ML1N is the cytoplasmic domain of TRPML1, which is described to bind PI(3,5)P₂ (Li et al., 2013). This plasmid was a generous gift from Haoxing Xu (Department of Molecular, Cellular, and Developmental Biology, University of Michigan, Ann Arbor, USA). Wildtype (WT) Atg5 and a dominant negative (DN) version were used to examine the impact of autophagosome formation. They were obtained from Noboru Mizushimathrough Addgene (#22949 and #22948). To follow particles positive for the autophagosome marker microtubule-associated protein 1A/1B-light chain 3 (LC3) we used RFP-EGFP tandem fluorescence fused to LC3, a gift from Tamotsu Yoshimori (Addgene plasmid #21074) as previously (Hilverling et al., 2021).

For immunofluorescence staining, neurons and HEK293T cells were fixed with 4% PFA and 5% sucrose, washed twice in PBS and once in 0.1% Triton X in PBS and incubated in blocking solution for 30 min (1% BSA and 0.1% Triton X in PBS). The first antibody against Lamp1 (Abcam ab 24170, rabbit polyclonal, 1:500), MAP2 (Abcam ab5392, chicken polyclonal, 1:2,000), or human α-synuclein (Enzo, ALX-804-258-L001, rat monoclonal, clone 15G7, 1:1,000) was incubated in 0.2% BSA in PBS overnight. For validation of the differentiation of LUHMES primary antibody against Ki-67 (Leica NCL-MM1, mouse, 1:500), Tyrosinhydroxylase (Chemicon, ab152, rabbit, 1:300), VMAT2 (Invitrogen, 48-0900, rabbit, 1:500), Nestin (Chemicon, abd69, rabbit 1:500) and beta III-Tubulin (Millipore, AB9354, chicken 1:500) was used. After washing, the secondary antibody tagged with Cy7 (Genecopeia, Rockville, USA, L144A, 1:600 in 0.2% BSA in PBS) or Alexa-647, Alexa-405, or Alexa-555 (Invitrogen Thermo Fisher Scientific, Waltham, USA, 1:500 in 0.2% BSA in PBS) was incubated at room temperature for 2 h in the dark. Coverslips were mounted on glass slides using Fluoromount-G (Southern Biotech, Birmingham, USA).

For the quantification of EGFP distribution pattern in HEK293T cells, the analysis was carried out as previously described (Dinter et al., 2016). Cells were transfected 24 h after plating on coverslips. Six hours after transfection, cells were chemically treated. Eighteen hours after treatment, cells were washed three times with cold phosphate-buffered saline (PBS) and fixed with paraformaldehyde (4% PFA, 5% sucrose in PBS) for 10 min. Subsequently, coverslips were washed three times in PBS and twice in ddH₂O. After letting them dry for half an hour, the coverslips were mounted with Fluoromont-G. Using fluorescence microscopy (Olympus BX51 microscope, Olympus, Hamburg, Germany, 40× oil objective), the EGFP distribution pattern was manually classified as homogenous distribution, particles (most likely aggregates), aggresome (large round aggregate), and unhealthy (round, condensed cells). At least 100 cells per coverslip were classified by an investigator blinded for the experimental condition. In each experiment, three coverslips were evaluated per group and the results averaged in mean. “n” corresponds to the number of independent experiments.

For analysis of the vesicle pools in HEK293Tcells, images were acquired with identical acquisition settings for all samples using fluorescence microscopy [Olympus IX81 equipped with 60× oil objective NA 1.3, Hammatsu CCD Camera, and Xcellence Software (Olympus, Hamburg, Germany)].

Images of NPC-derived neurons and LUHMES-derived dopaminergic neurons were acquired with identical acquisition settings using confocal microscopy [Zeiss—Xio Observer.Z1 with inverted stand, 40× air objective, Yokogawa CSU-X1M 5000 dual camera and ZEN Blue 2011 Software (Zeiss, Oberkochen, Germany)]. Images were processed in ImageJ (Version 2.1.0/1.53c) and particle appearance was classified manually blinded for the experimental group. In total, we analyzed 31 NPC-derived neurons and 30 LUHMES-derived dopaminergic neurons for DMSO control and 30 NPC-derived neurons, and 30 LUHMES-derived dopaminergic neurons with ML-SA1 treatment.

To determine the position and abundance of particles within HEK293T cells, a “shrink analysis” was used as previously described (Hilverling et al., 2021). Briefly, the cell was outlined as a freehand region of interest (ROI). Using the published custom-written macro the cell outline was progressively eroded by 10% of its original size until only 10% of the original size was left. In each step, the number of particles in the entire cell and the remaining area was quantified.

For time-lapse microscopy, cells were grown in 24-well plates with glass bottom (Greiner bio-one, Frickenhausen, Germany) and transfected as described above. One hour before imaging, the medium was changed to Hank’s Buffered Salt Solution with 100 nM BafA1 to induce autophagy by starvation and to block the final steps of autophagy. MLSA1 and DMSO were added directly before starting imaging. We used a Zeiss spinning disk confocal microscope (Zeiss blue software, 40× air objective) equipped with an incubator (37°C, 5% CO₂) and a motor stage to acquire images at defined positions every 20 min over 80 min. Changes in particle appearance were classified manually based on the resulting images by an investigator blinded for the experimental group. Analysis was done in ImageJ. In total, we analyzed 31 cells for DMSO control and 32 cells with ML-SA1 treatment.

To determine the highest non-toxic concentration of ML-SA1 and ML-SI3 on LUHMES cells, the concentration of extracellular LDH was measured in the medium 24 h after the last medium change with the start of chemical treatment on day 9 of differentiation using the LDH-Cytox Assay Kit (Biolegend, San Diego, California, USA) according to the manufacturer’s instruction. Briefly, for each independent condition, three to four technical replicates were measured. Absorbances were measured at 492 nm; the reference wavelength was 620 nm. Background (medium) absorbance was subtracted from all values, and values were normalized to control cells (DMSO-treated) and to maximal lysed cells (treated with 1% Triton X-100 for 10 min).

Immunoblots for HEK cell lysates were carried out as previously described (Dinter et al., 2016). HEK293T cells were plated in a 6-well plate, transfected, and treated as described above. Twenty-four hours after transfection, cells were washed three times with PBS and lysed for 30 min at 4°C in NP40-lysis buffer (0.5% NP40, 50 mM Tris, 100 mM NaCl, 5 mM MgCl₂, 1 mM EDTA, pH 8.8) containing protease inhibitor (Pierce, Rockford, IL, USA, Thermo Fisher Scientific, Rockford, USA). The total protein concentration was determined using the DC Protein Assay Kit (Bio-Rad, Munich, Germany). Twenty microgram of protein was subjected to 12% SDS polyacrylamide gel electrophoresis and proteins were blotted onto nitrocellulose membranes. Primary antibodies were incubated overnight at 4°C and HRP-conjugated secondary antibody was incubated for 2 h at room temperature. Detection was carried out by chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate, Thermo Fisher Scientific, Rockford, USA). Bands were quantified using the software BioDocAnalyze (Biometra, Göttingen, Germany) as previously (Dinter et al., 2016). The following antibodies were used: rabbit anti-α-synuclein (1:500, #2642, Cell Signaling Technology, Danvers, USA), rabbit anti-actin (1:1,000, abcam, ab1801, Cambridge, UK), and anti-rabbit IgG (1:10,000, GE Healthcare Life science, Freiburg, Germany).

As described previously, incubation with the α-synuclein antibody in HEK293T lysates showed an unspecific band at around 35 kDa and two bands around 20 kDa. The upper one was only seen in cells transfected with α-synuclein and, therefore, considered as the α-synuclein band (Dinter et al., 2016).

Immunoblot analysis for LUHMES cells and hiPSC-derived neuron lysates was performed as described before (Szego et al., 2019). For protein quantification, cells were lysed in a buffer containing 1% Triton X-100, 25 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, and protease inhibitors. After centrifugation (13,000 g, 20 min, 4°C). Twenty microgram of Triton X-100 soluble lysate was loaded onto a 12% Bis-Tris gel for Western blot analysis (NuPAGE, ThermoFisher, USA). Every lysate was run twice in parallel for dry blotting on nitrocellulose and PVDF-membrane in parallel (iBlot, ThermoFisher, USA). After blocking, membranes were incubated with first antibodies against human αSyn (1:4,000, rat, Enzo, ALX-804-258-L001), LC3 (1:2,000, rabbit, Cell Signaling, #2775), p62 (1:5,000, mouse, Abcam, ab56416) and beta-tubulin (1:3,000, rabbit, Cell Signaling, #2146S), HRP-conjugated secondary antibodies were incubated for 1.5 h at room temperature. The signal was visualized with chemiluminescent substrate (SuperSignal West Dura/Femto Chemiluminescent Substrate, Thermo Fisher Scientific, Rockford, USA) and detected with ImageQuant LAS 4000.

For statistical analysis and presentation of data, GraphPad Prism 7.0 was used (GraphPad Software, La Jolla, USA). Graphs represent mean ± SEM. The test used for comparison in each graph and the number of independent experiments are noted in the figure legend. p < 0.05 was considered statistically significant. p values are depicted as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Non-significant differences are depicted as n.s.

To confirm that activating TRPML-1 affects α-synuclein pathology in human cells, we expressed the pathogenic A53T mutant of α-synuclein in HEK293T cells. A53T-α-synuclein was flexibly tagged with EGFP by the interaction of a six amino acid PDZ binding motif added to the C-terminus of α-synuclein, as described (Opazo et al., 2008). Using this technique, aggresomes, aggregates, and vesicular α-synuclein particles can be detected in living and fixed cells. Cells were manually classified regarding their EGFP distribution patterns (Supplementary Figure 1A) cells with a homogenous distribution of EGFP (“homogenous”), cells showing a punctate distribution of EGFP—most likely aggregates (“aggregates”), cells showing round, large aggregates (“aggresome”) and cells with a rounded appearance (“unhealthy”), which suffer apoptosis when followed by time-lapse microscopy. This cell classification assay has been used in previous studies to determine differences between α-synuclein variants and the effects of therapeutic manipulations (Opazo et al., 2008; Karpinar et al., 2009; Krumova et al., 2011; Dinter et al., 2016; Saridaki et al., 2018).

To activate TRPML1, we used its small molecule activator ML-SA1 in a final concentration of 25 μM. In the control group, 46% of cells showed a homogenous distribution of EGFP, i.e., no number of α-synuclein particles (Figures 1A–C). Application of ML-SA1 increased the number of homogenous cells significantly to 60% and reduced the percentage of cells with punctate distribution (Figures 1B–D). The average number of cells with an aggresome and with round unhealthy appearance were smaller with ML-SA1 treatment, but this difference was not statistically significant—probably due to the small number of cells in these categories (Figures 1E,F).

To determine whether activation of TRPML-1 also reduces the total amount of α-synuclein protein, we performed immunoblot analysis of HEK293T cells transfected and treated as described above. The specific α-synuclein immunoreactivity decreased by 23% under treatment with ML-SA1 compared to the control (Figures 1G,H).

Phosphoinositides (PIs) play a key role in various cellular processes involving signaling (Di Paolo and De Camilli, 2006; Ariosa and Klionsky, 2016). Activation of TRPML-1, for instance, requires PI(3,5)P₂ (Dong et al., 2010; Fine et al., 2018). PI(3,5)P₂ is generated from PI3P by phosphorylation at the 5-position of the inositol ring by PIKfyve. To confirm that ML-SA1 indeed acts throughTRPML1, we used a protein that binds and sequesters PI(3,5)P₂, ML1N*2, the N-terminal cytoplasmatic domain of TRPML1 (Dong et al., 2010; Li et al., 2013). We co-expresssed ML1N*2 with A53T-α-synuclein and classified the EGFP distribution pattern upon ML-SA1 treatment. Indeed, coexpression of ML1N*2 blocked the ML-SA1 effect (Figure 2A, only the numbers for cells with homogenous distribution patterns are shown for clarity). Interestingly, expression of ML1N*2 increased the baseline percentage of cells with a homogenous distribution of EGFP.

To verify this finding, we blocked the synthesis of PI(3,5)P₂ using the PIKfyve inhibitor YM201636. Under control conditions, ML-SA1 increased the percentage of cells with a homogenous distribution. This effect was blocked by treatment with YM201636 (Figure 2B). Together with the effect of ML1N*2, this finding indicates that ML-SA1 indeed acts through the PI(3,5)P₂-dependent TRPML1 channel.

TRPML1 is a cation channel and likely mediates the release of Ca²⁺ from lysosomes into the cytosol (Dong et al., 2010). We, therefore, hypothesized that a rise of cytosolic Ca²⁺might mediate the ML-SA1 effect on α-synuclein aggregates. To test this hypothesis, we depleted cytosolic Ca²⁺ using the cell-permanent calcium chelator BAPTA. As Ca²⁺ is essential for many cellular functions, we first determined the highest non-toxic concentration of BAPTA-AM. 100 nM final concentration of BAPTA-AM was associated with a similar number of dead cells compared to untreated control (Supplementary Figure 1B). At this concentration, prolonged exposure to BAPTA-AM did not alter the number of cells with a homogenous distribution of EGFP-tagged A53T-α-synuclein (Figure 2C). The effect of ML-SA1, in contrast, was not observed in the presence of 100 nM BAPTA-AM (Figure 2C). This result indicates that a rise in cytosolic calcium is involved in the effect of ML-SA1 on α-synuclein aggregates.

The ML-SA1 induced decrease of α-synuclein particles and protein levels can be explained by increased clearance. Autophagy—where cytosolic content gets packed into vesicles—is the main degradation pathway of α-synuclein. To discriminate between cytosolic α-synuclein aggregates and α-synuclein aggregates engulfed by a membrane, we used a chemically induced dimerization approach, as described before (Banaszynski et al., 2005; Dinter et al., 2016; Hilverling et al., 2021). We transfected HEK293T cells with the dimerization domain FK506 binding protein (FKBP) fused to EGFP and in addition with mCherry-A53T-α-synuclein fused to the other dimerization domain FKBP-rapamycin binding domain (FRB). Six hours after transfection, cells were treated with ML-SA1 or DMSO for 18 h. One hour before fixation, cells were additionally treated with rapamycin to induce the hetero-dimerization of EGFP-FKBP and A53T-α-synuclein-mCherry-FRB (Figure 3). Colocalization of GFP and mCherry only occurs for cytosolic α-synuclein particles (black arrows in Figures 3A,B) whereas A53T-α-synuclein-mCherry-FRB contained in vesicles is not accessible to EGFP-tagged FKBP (white arrows in Figures 3A,B).

In control cells, around three α-synuclein particles per cell were classified as inside vesicles (Figures 3A–C). Upon ML-SA1 treatment, the number of α-synuclein particles classified as vesicles increased threefold (Figures 3B,C). ML-SA1 thus increased the number of vesicles containing α-synuclein, which most likely represent autophagosomes.

Since TRPML1 can regulate lysosomal mobility, we also analyzed the subcellular location of α-synuclein-positive particles. There was no significant effect of ML-SA1 on the subcellular location of cytosolic particles and vesicles (Supplementary Figures 2A,B).

Given that ML-SA1 increased the number of α-synuclein-containing vesicles—most probably autophagosomes—we took a deeper look into the different steps along the autophagy pathway.

To investigate whether activation of TRPML1 by ML-SA1 induces de novo formation of autophagosomes in our paradigm, we expressed the autophagosomal marker LC3-EGFP and treated the cells with ML-SA1 or DMSO as a control. Cells were manually classified as cells with a homogenous distribution pattern of LC3-EGFP and cells with LC3-EGFP-positive puncta (Figure 4A). Visible puncta of LC3 are only formed when LC3 accumulates on early autophagosomal membranes; LC3-positive puncta are thus defined as autophagosomes (Pankiv et al., 2007). Interestingly, administration of ML-SA1 for 18 h increased the percentage of cells with a homogenous distribution (Figures 4A,B). The number of cells with few LC3-GFP puncta was not altered by ML-SA1, the number of cells with many puncta was decreased. Scotto Rosato et al. (2019) described the biogenesis of early autophagic vesicles by ML-SA1 and other TRPML1 agonists that could be observed already a few minutes after application. Our finding, therefore, suggests that enhancing the biogenesis of early autophagic vesicles is not the only effect of ML-SA1. Rather, ML-SA1 might also affect the late steps of autophagy, increasing autophagosome maturation and clearance as suggested by the lower number of cells with LC3-tagged autophagosomes (Figure 4B).

In order to test whether the first steps of autophagy are at all required for A53T-α-synuclein particle reduction by ML-SA1, we first used 3-Methyladenine (3-MA), which inhibits autophagy initiation (Blommaart et al., 1997). The reduction of A53T-α-synuclein particles by ML-SA1 was not observed in the presence of 3-MA (Figure 4C). To confirm this finding by genetic manipulation, we expressed WT and DN autophagy-related 5 (Atg5). Atg5 plays a crucial role in the first steps of autophagy, specifically in the formation of the autophagosomal membrane (Hurley and Young, 2017). Indeed, the effect of ML-SA1 treatment was not observed in cells expressing DN Atg5, but in cells expressing WT Atg5 (Figure 4D). Conversely, when autophagy was initiated by treatment with rapamycin, the number of cells without α-synuclein increased, but ML-SA1 had no additional effect (Figure 4F). Taken together, these findings indicate that ML-SA1 affects A53T-α-synuclein particles through autophagy.

Since ML-SA1 might affect the late steps of the autophagy-lysosome pathway, we next tested the effect of Bafilomycin A1 (BafA1), which blocks the fusion of autophagosomes with lysosomes. This effect most probably results from a block of the sarco-endoplasmatic calcium ATPase. In addition, BafA1 inhibits acidification of lysosomes and endosomes by altered H⁺-ATPase activity (Mauvezin and Neufeld, 2015; Klionsky et al., 2016). Treatment with BafA1 decreased the percentage of cells with a homogenous distribution of A53T-α-synuclein tagged with EGFP (black bars in Figure 4E), consistent with clearance of A53T-α-synuclein particles by the autophagy-lysosome pathway. ML-SA1 still showed an effect in cells treated with BafA1 (Figure 4E). This result indicates that ML-SA1 may be able to override the effect of BafA1 on autophagosome-lysosome fusion and/or acidification. Taken together, these findings indicate that ML-SA1 likely acts on α-synuclein at the late steps of the autophagy-lysosome-pathway.

In order to explore whether ML-SA1 affects the abundance of acidic autophagosomes in general, i.e., without α-synuclein, HEK293T cells were transfected with the autophagosomal marker LC3 tagged with tandem fluorescent tag EGFP-mRFP (TFL) as described before (Dinter et al., 2016; Hilverling et al., 2021). In short, TFL allows a distinction between vesicles with acidic and neutral luminal pH. In neutral pH, GFP and mRFP are functional and appear yellow in merged images. EGFP is quenched by acidic pH, so acidic vesicles appear red in merged images. After ML-SA1 treatment for 24 h, we observed fewer neutral LC3-positive vesicles (neutral/early autophagosomes) and a stable population of acidic LC3-positive vesicles (mature autophagosomes; Figure 5A). The fraction of acidic vesicles was thus increased (Figure 5B). This finding could be explained by ML-SA1 promoting the maturation of autophagosomes. Hence, mature autophagosomes could be cleared during the 24 h incubation.

To observe the immediate effects of ML-SA1 on neutral and acidic autophagosomes, we performed time-lapse imaging of HEK293T cells expressing LC3-TFL. To obtain a sufficient number of autophagosomes for imaging, we induced autophagy by starvation and blocked autophagosome clearance by treatment with BafA1. Consequently, the number of neutral autophagosomes showed a continual increase over the time of recording (Figure 5C black). As expected, the number of mature autolysosomes did not significantly change over time in the presence of BafA1 (Figure 5D black). With MLSA1, the continual increase in neutral autophagosomes over time was significantly less pronounced after 60 min compared to the control condition (Figure 5C orange). This finding is consistent with the data at 24 h obtained without starvation and BafA1 (Figure 5A). It could be explained by the increased maturation of neutral autophagosomes into acidic autophagosomes by ML-SA1. Accordingly, the number of mature autolysosomes increased over time (Figure 5D orange). In total, the percentage of acidic vesicles after 80 min was significantly higher with ML-SA1 compared to control (Figure 5E), again consistent with the findings after 24 h (Figure 5B). This finding suggests that ML-SA1 can partially override the BafA1 effect, consistent with the observation in Figure 4E. The increase in acidic LC3 vesicles was already observed after 20 min, consistent with the rapid onset observed by others (Scotto Rosato et al., 2019).

As the final steps in the autophagy-lysosome pathway might be crucial for the effect of ML-SA1 on α-synuclein, we took a closer look into the effect of ML-SA1 on the different maturation steps during autophagy. Using TFL-tagged α-synuclein and the lysosomal marker Lamp1 as described before (Hilverling et al., 2021), the vesicles involved in these steps can be classified according to their luminal pH and the composition of their membranes. We distinguished five different vesicle types (Figure 6A). ML-SA1 significantly reduced the number of Lamp1-negative neutral α-synuclein particles, which are considered cytosolic aggregates or neutral autophagosomes (Figure 6B). This effect of ML-SA1 was not observed when cells were incubated with BafA1, indicating that fusion of autophagosomes with lysosomes or vesicle acidification is required for the reduction of α-synuclein aggregates and neutral autophagosomes by ML-SA1.

ML-SA1 did not change the abundance of early immature autolysosomes, which are already Lamp1-positive but still neutral vesicles (Figure 6C). BafA1 strongly reduced this population (Figure 6B), consistent with its known inhibition of autophagosome-lysosome fusion. This reduction was partially rescued by ML-SA1, in accordance with the hypothesis that ML-SA1 could partially override the blockade of autophagosome-lysosome fusion by BafA1 noted above (Figures 4E, Figure 5D). ML-SA1 increased the number of mature autolysosomes, which are Lamp1-positive and acidic vesicles (Figure 6D). This effect was blocked by BafA1, suggesting that ML-SA1 might be better at overriding the inhibition of autolysosomal fusion by BafA1 than at rescuing vesicle acidification. ML-SA1 also increased the number of amphisomes, Lamp1-negative acidic vesicles (LNAV; Figure 6E). This effect was also blocked by BafA1. Finally, ML-SA1 increased the number of lysosomes, i.e., Lamp1-positive vesicles that are not positive for α-synuclein (Figure 6F). This effect was not observed in the presence of BafA1 (Figure 6F). Mature autolysosomes in which the TFL-tag has been degraded could be misclassified as lysosomes, which can potentially explain this effect. In addition, the increased abundance of lysosomes with ML-SA1 can be explained by autolysosome reformation, i.e., the fact that new lysosomes are formed from autolysosomes (Yu et al., 2010; Li et al., 2016). Autolysosome reformation could be stimulated directly by ML-SA1, or potentiated by the increased abundance of autolysosomes (Figure 6D).

TRPML is also known for regulating the intracellular transport of lysosomal vesicles to the cell periphery after starvation (Li et al., 2016). To analyze the possible effects of ML-SA1 on vesicle localization, we performed an analysis of the subcellular distribution of all five particle types (Supplementary Figure 2) as described (Hilverling et al., 2021). We did not observe a significant effect of ML-SA1 on the vesicle localization.

Taken together, these findings with α-synuclein-tagged particles confirm our results with LC3-tagged vesicles (Figure 4) that ML-SA1 reduces the abundance of neutral particles (Figure 5B) and increases the abundance of autolysosomes (Figures 5C,D potentially) and acidic vesicles (Figures 5D,E). The effect of ML-SA1 on the subcellular localization of autophagolysosomal vesicles was not pronounced.

Inhibition of mTOR induces autophagy but blocks autolysosome reformation (Yu et al., 2010; Zhang et al., 2016). In order to test the involvement of autolysosome reformation for the ML-SA1 effect, we tested the effect of the mTOR inhibitor rapamycin. As mentioned above, ML-SA1 had no additional effect on EGFP-tagged A53T-α-synuclein in the presence of rapamycin (Figure 4F). Moreover, rapamycin inhibited the effect of ML-SA1 on immature and mature autolysosomes as reported by TFL-tagged A53T-α-synuclein and Lamp1 staining (Figures 5C,D). In particular, the number of lysosomes increased with ML-SA1 in the control condition but decreased with ML-SA1 in the presence of rapamycin (Figure 5E). As one possible explanation, the induction of autophagosome-lysosome fusion by ML-SA1 could deplete lysosomes, in cases where lysosome reformation is inhibited by rapamycin.

In order to validate the findings in a neuronal cell type, we analyzed the occurrence of Lamp1-positive and Lamp1-negative α-synuclein particles in mature dopaminergic neurons derived from LUHMES cells, which are immortalized human dopaminergic midbrain neurons (Scholz et al., 2011; Neuhof et al., 2021). WT-α-synuclein and GFP were expressed by lentiviral transduction of an aggregate-prone T2A-based construct (Szego et al., 2019). After 1 week of expression, neurons were exposed to ML-SA1 or DMSO for 24 h, its non-toxic concentration was defined by LDH-assay (Supplementary Figure 4C). We stained against human α-synuclein (gray in Figures 7A,B), the lysosomal marker Lamp1 (red in Figures 7A,B), and the neuronal marker MAP2 (blue in Figure 7A). We observed several puncta that were positive for α-synuclein and Lamp1 (e.g., closed arrowheads in Figure 7B), but also puncta that were only positive for synuclein (e.g., open arrowheads in Figure 7B) and puncta only positive for Lamp1 (e.g., asterisk in Figure 7B). We hypothesize that these categories represent autolysosomes, autophagosomes (or cytosolic aggregates), and lysosomes. Neurons treated with ML-SA1 showed a significantly higher percentage of α-synuclein in autolysosomes than in the control condition (Figure 7C).

ML-SI3 is a new compound, found to be an antagonist to TRPML1, but also affects TRPML2 and 3 to some extent (Leser et al., 2021). In order to validate our findings on autophagosome formation (Figures 4, 5) using a different technique, we quantified in lysates of dopaminergic neurons derived from LUHMES cells the fraction of lipidated LC3 (LC3II/LC3I ratio) using immunoblots. Consistent with our findings in HEK293T cells, ML-SA1 and ML-SI3 did not change the amount of lipidated LC3 after 2 h (Figures 7D,E) or 24 h (Supplementary Figures 5A,B). p62 is an autophagic cargo commonly used to quantify autophagic flux (Klionsky et al., 2016). p62 significantly accumulated in the presence of the TRPML inhibitor ML-SI3 (Figures 7F,G), whereas ML-SA1 showed a trend for reduced p62 after 24 h (Figures 7F,G). Yet, administration of ML-SA1 and ML-SI3 for 24 h did not show a significant effect on the amount of overexpressed α-synuclein analyzed in LUHMES cells 9 days after lentivirus-based transduction with α-synuclein-T2A-GFP (Figures 7F–H).

In order to check the effect of ML-SA1 and ML-SI3 on α-synuclein in more detail, we used mature neurons that were derived from human-induced pluripotent stem cells (hiPSC). Neurons were analyzed 21 days after lentivirus-based transduction with α-synuclein-T2A-GFP (Figure 8A). By staining for Lamp1, we observed with ML-SA1 a significantly higher percentage of α-synuclein that was packed into autolysosomes (Figures 8B,C). In immunoblot analysis upon ML-SA1 or ML-SI3 treatment, the fraction of lipidated LC3 were not altered (Supplementary Figure 6C). Both findings are consistent with our findings in LUHMES-derived dopaminergic neurons (Figures 7C–E). With respect to α-synuclein, however, we observed a significant increase of aggregate-prone α-synuclein-T2A-dimers with the TRPML1 inhibitor ML-SI3 (Figures 8D,E). The agonist ML-SA1 showed an opposite trend that was not statistically significant. BafA1 increased the total amount of α-synuclein in neurons, and this accumulation was reversed by the TRPML agonist ML-SA1 (Figures 8D–F). The potentiated effect of ML-SA1 in the presence of BafA1 is consistent with TRPML1-induced lysosomal exocytosis of α-synuclein in neurons with inhibited lysosomal degradation (Tsunemi et al., 2019), but also with our findings of enhanced maturation of autophagosomes in the presence of BafA1 in HEK293T cells (Figure 5). Overall, these findings confirm in neurons that TRPML1 promotes autophagosome maturation and increases the abundance of α-synuclein-containing autolysosomal vesicles.

ML-SA1 reduced the amount of α-synuclein protein in HEK293T cells (Figure 1H) and the number of visible α-synuclein particles (Figure 1D). This effect was not observed when the initiation of autophagy was blocked by 3-MA or DN Atg5 (Figures 4C,D), blunted when late stages of autophagy were blocked by BafA1 (Figure 4E), and also not observed in the presence of rapamycin (Figure 4F). Of note, rapamycin has diverse cellular effects. The induction of autophagy and the activation of TRPML (Zhang et al., 2019) could contribute to the potential ceiling effect in Figure 4F and the potential floor effect in Figure 5B whereas reduced lysosome reformation (Yu et al., 2010) best explains the effect on lysosomes in Figure 5F. Collectively, the cited findings confirm that ML-SA1 reduces α-synuclein through the auto-lysosomal pathway and not through other mechanisms of α-synuclein clearance such as the proteasome. Accordingly, α-synuclein was observed less frequently in free cytosolic aggregates and more commonly inside vesicles with ML-SA1 in HEK293T cells (Figure 3D). Moreover, α-synuclein was more commonly colocalized in autolysosomes with ML-SA1 in human dopaminergic neurons (Figure 7C) and iPS cell-derived neurons (Figure 8C). The observed reduction of α-synuclein load in HEK293T cells is consistent with the reduction of α-synuclein by ML-SA1 via enhanced lysosomal exocytosis in stem cell derived neurons recently observed by others (Tsunemi et al., 2019) and with a large body of literature demonstrating accumulation of α-synuclein with autophago-lysosomal impairment, for instance, caused by the mutation in GBA or deficiency in ATP13A2 (Mazzulli et al., 2011; Usenovic et al., 2012; Schultheis et al., 2013; Lopes da Fonseca et al., 2016). Accordingly, enhanced α-synuclein clearance can be reached by facilitating auto-lysosomal function, for instance by increasing GBA or Rab7 (Dinter et al., 2016; Mazzulli et al., 2016; Saridaki et al., 2018; Szegö et al., 2022).

We assume that ML-SA1 achieved these effects by activation of the TRPML1 cation channel because the effects were not observed in the presence of the calcium chelator BAPTA (Figure 2C) and required the phosphoinositide PI(3,5)P₂ (Figures 2A,B) that TRPPML1 function depends on (Dong et al., 2010). We note that the ML1N construct has further binding partners than PI(3,5)P₂ (Hammond et al., 2015) that could underlie the altered baseline with ML1N (Figure 2A), but the lacking effect of ML-SA1 in the presence of ML1N is consistent with ML-SA1 acting through TRPML1. Overall, these findings are consistent with the fact that TRPML mediates calcium release from lysosomes (Grimm et al., 2012; Shen et al., 2012), with the dependence of lysosomal degradation of endosomal cargo on PIKfyve (de Lartigue et al., 2009) and the accumulation of autophagosomal cargo with blockade of Figure 4 and Vac14, which also regulate PI(3,5)P₂ (Ferguson et al., 2009).

TRPML1 functions in different lysosome-dependent compartments, including autophagosomes, mitochondrial membrane contact sides, lysosome transport, and lysosome exocytosis (Scotto Rosato et al., 2019; Tsunemi et al., 2019; Peng et al., 2020). Accordingly, different mechanisms have been proposed through which TRPML1 agonists like ML-SA1 produce their effects. For clarity, we discuss here separately the effects of TRPML1 agonists in cells with intact and impaired auto-lysosomal function.

Studying cells with an intact auto-lysosomal system is difficult because there are only a few autophagosomes (Figure 4A). Moreover, the abundance of any vesicle subtype results from a steady state of formation and degradation processes. Autophagosomes are rapidly processed along their maturation pathway, i.e., they fuse with lysosome, acidify, and are reformed into lysosomes. Accordingly, the number of mature autolysosomes—as reported by the TFL tag—is even lower than the number of neutral autophagosomes (Figure 6A). In our hands, 18 h of ML-SA1 increased the number of cells without visible autophagosomes and decreased the number of cells with many autophagosomes (Figure 4B). In particular, the number of neutral autophagosomes decreased (Figure 6B). These findings are best explained by improved autophagosome maturation with ML-SA1 and cannot be explained by the transcriptional activation by TRPML1 of autophagic and lysosomal genes through the transcription factor TFEB (Medina et al., 2015). Accordingly, we did not find an effect of ML-SA1 or ML-SI3 on LC3 lipidation in neurons (Figure 7E, Supplementary Figure 6C, but increased colocalization of α-synuclein with Lamp1; Figures 7C, 8C).

In order to increase the abundance for autophagosomes and reduce their rapid maturation, we used time-lapse microscopy with LC3-TFL cells that were starved and incubated with BafA1. In this paradigm, we observed a continual increase of neutral autophagosomes (Figure 5C) as expected with impaired auto-lysosomal fusion and acidification. ML-SA1 led to an increase in the number of acidic vesicles and to a less pronounced accumulation of neutral autophagosomes (Figures 5C,D). This observation is best explained by improved maturation and thus acidification of autophagosomes, resulting in a reduced accumulation of neutral autophagosomes. The fact that the increase in acidic autolysosomes was transient can be explained by the dwindling number of available neutral precursors and by the fact that degradation of LC3-TFL protein in acidic autolysosomes remained intact.

Based on the TFL-tag alone we cannot discriminate whether ML-SA1 primarily increased autolysosomal fusion or vesicle acidification. Increased autolysosome fusion is consistent with the higher number of vesicles double-stained for LC3 and Lamp1 observed by others (Scotto Rosato et al., 2019) and with the dependence of ML-SA1 on a cytosolic calcium release (Figure 2C), which is a typical trigger for vesicle fusion events mediated by SNARE proteins. For instance, calcium release can promote th fusion of lysosomes with endosomes (Cao et al., 2015) or the plasma membrane (Tsunemi et al., 2019). Depletion of calcium stores by thapsigargin, on the other hand, blocks the fusion of autophagosomes with lysosomes (Ganley et al., 2011; Mauvezin and Neufeld, 2015; Mauvezin et al., 2015). Dependence on calcium release is also consistent with the rapid onset of the ML-SA1 effect observed in time-lapse microscopy (Figure 5D).

The effects of improving auto-lysosomal clearance are best observed in a context with impaired auto-lysosomal clearance. Indeed, TRPML1 agonists have been explored as potential therapeutic strategies to overcome auto-lysosomal dysfunction (Chen et al., 2014; Tedeschi et al., 2019; Tsunemi et al., 2019).

In this work, we have focused on the effect of the TRPML agonist ML-SA1 on clearance of α-synuclein in human cells, including dopaminergic neurons and stem cell derived neurons. α-synuclein inhibits autophagosomal turnover by impairing SNAP-29-dependent auto-lysosomal fusion (Tang et al., 2021). In HEK293T cells expressing A53T-α-synuclein, ML-SA1 reduced the abundance of cytosolic α-synuclein aggregates and increased the abundance of α-synuclein in vesicles (Figure 3D). This shift is best explained by increased autophagy of α-synuclein aggregates and is consistent with the observation of increased autophagosome biogenesis with TRPML1 agonists in starved HeLa cells (Scotto Rosato et al., 2019). Yet, increased autophagosome biogenesis cannot explain all the effects of ML-SA1.

In neurons with loss of function mutation in ATP13A2/PARK9, which disrupts lysosomal exocytosis and leads to accumulation of α-synuclein (Tsunemi and Krainc, 2014; Tsunemi et al., 2014), ML-SA1 reduces α-synuclein primarily by facilitating fusion of lysosomes with the plasma membrane, i.e., α-synuclein exocytosis (Tsunemi et al., 2019). We did not assess the effect of ML-SA1 on α-synuclein exocytosis in this study, but enhanced α-synuclein exocytosis cannot explain the higher number of α-synuclein-containing vesicles (Figure 3D), and the shift in vesicle populations (Figure 6).

In our paradigm, ML-SA1 was able to override the block of autolysosome fusion and acidification by BafA1, both in HEK293T cells and in neurons (Figures 4E, 5E, 6, 8H), suggesting that ML-SA1 affects the steps in autophagosome maturation inhibited by BafA1. Accordingly, ML-SA1 produced a shift of α-synuclein-tagged vesicles towards Lamp1-positive vesicles and towards acidic compartments (Figures 5D–F, 7C, 8C). The increase of immature autolysosomes with ML-SA1 in the presence of BafA1 (Figure 6C) indicates that ML-SA stimulates autolysosome fusion, consistent with TRPML being a calcium channel as discussed above. Yet, intracellular calcium not only mediates vesicle fusion but has also been implicated in acidification (Luzio et al., 2007; Morgan et al., 2011). The higher abundance of mature autolysosomes but not immature autolysosomes with ML-SA1 (Figures 6C,D) can be interpreted as evidence for an effect of ML-SA1 on vesicle acidification, but could also result from the abundance of immature autolysosomes being a steady-state of several competing processes. Consistently, phagosome acidification was recently shown to depend on TRPML1 and to be rescued by ML-SA1 in PIKfyve deficient cells (Isobe et al., 2019).