Advertisement

High-content phenotypic screen to identify small molecule enhancers of Parkin-dependent ubiquitination and mitophagy

Open AccessPublished:January 03, 2023DOI:https://doi.org/10.1016/j.slasd.2022.12.004

      Highlights

      • p-Ser65-ubiquitin is a specific biomarker of PINK1-Parkin-dependent mitophagy initiation.
      • High-content phenotypic screening (HCS) identified compounds increasing p-Ser65-ubiquitin in Parkinson's disease patient-derived fibroblasts.
      • HCS-derived hit compounds were triaged using a panel of counter-screening assays.
      • Hit compounds enhance downstream mitochondrial clearance, a key functional mitophagy endpoint.
      • Mechanism-of-action studies of HCS-derived hit compounds identified inhibitors of USP30, a negative regulator of mitophagy.

      Abstract

      Mitochondrial dysfunction and aberrant mitochondrial homeostasis are key aspects of Parkinson's disease (PD) pathophysiology. Mutations in PINK1 and Parkin proteins lead to autosomal recessive PD, suggesting that defective mitochondrial clearance via mitophagy is key in PD etiology. Accelerating the identification and/or removal of dysfunctional mitochondria could therefore provide a disease-modifying approach to treatment. To that end, we performed a high-content phenotypic screen (HCS) of ∼125,000 small molecules to identify compounds that positively modulate mitochondrial accumulation of the PINK1-Parkin-dependent mitophagy initiation marker p-Ser65-Ub in Parkin haploinsufficiency (Parkin +/R275W) human fibroblasts. Following confirmatory counter-screening and orthogonal assays, we selected compounds of interest that enhance mitophagy-related biochemical and functional endpoints in patient-derived fibroblasts. Identification of inhibitors of the ubiquitin-specific peptidase and negative regulator of mitophagy USP30 within our hits further validated our approach. The compounds identified in this work provide a novel starting point for further investigation and optimization.

      Graphical abstract

      Keywords

      Abbreviations:

      AA/O (antimycin A/oligomycin A), DUB (deubiquitinase), FAC (final assay concentration), HCS (high-content screen), IMM (inner mitochondrial membrane), MMP (mitochondrial membrane potential, OMM, outer mitochondrial membrane), PD (Parkinson's disease), PPA (putative Parkin activator), TMRM (tetramethyl rhodamine methyl ester), USP (ubiquitin specific peptidase)

      1. Introduction

      Mitochondrial dysfunction is a prominent pathological feature of both sporadic and familial Parkinson's disease (PD) and has long been implicated in disease etiology [
      • Osellame L.D.
      • et al.
      Mitochondria and quality control defects in a mouse model of Gaucher disease-links to Parkinson's disease.
      ,
      • Valente E.M.
      • et al.
      Hereditary early-onset Parkinson's disease caused by mutations in PINK1.
      ,
      • Clark I.E.
      • et al.
      Drosophila pink1 is required for mitochondrial function and interacts genetically with Parkin.
      ,
      • Park J.
      • et al.
      Mitochondrial dysfunction in Drosophila PINK1 mutants is complemented by Parkin.
      ,
      • Kitada T.
      • et al.
      Mutations in the parkin gene cause autosomal recessive juvenile parkinsonism.
      ,
      • Fiesel F.C.
      • et al.
      Patho-)physiological relevance of PINK1-dependent ubiquitin phosphorylation.
      ,
      • Hou X.
      • et al.
      Age- and disease-dependent increase of the mitophagy marker phospho-ubiquitin in normal aging and Lewy body disease.
      ,
      • Silvestri L.
      • et al.
      Mitochondrial import and enzymatic activity of PINK1 mutants associated to recessive parkinsonism.
      ]. Monogenic PD implicates mitochondria as central to disease pathogenesis [
      • Clark E.H.
      • Vázquez de la Torre A.
      • Hoshikawa T.
      • Briston T.
      Targeting mitophagy in Parkinson's disease.
      ]. Mutations in mitochondrial PTEN (phosphatase and tensin homologue)-induced kinase 1 (PINK1; encoded by PARK6) and Parkin (encoded by PARK2/PRKN), cause autosomal recessive early-onset PD [
      • Valente E.M.
      • et al.
      Hereditary early-onset Parkinson's disease caused by mutations in PINK1.
      ,
      • Kitada T.
      • et al.
      Mutations in the parkin gene cause autosomal recessive juvenile parkinsonism.
      ,
      • Silvestri L.
      • et al.
      Mitochondrial import and enzymatic activity of PINK1 mutants associated to recessive parkinsonism.
      ,

      Padmanabhan S., Polinski N.K., Menalled L.B., Baptista M.A.S., Fiske B.K. The Michael J. Fox Foundation for Parkinson’s research strategy to advance therapeutic development of PINK1 and Parkin. Biomolecules 2019 Jul 24;9(8):296. doi:10.3390/biom9080296.

      ,
      • Lesage S.
      • et al.
      Characterization of recessive Parkinson disease in a large multicenter study.
      ]. PINK1 and Parkin act in concert to identify and subsequently remove damaged mitochondria by selective autophagy (mitophagy), being, respectively, sensor and amplifier proteins of the mitophagy process [
      • Clark I.E.
      • et al.
      Drosophila pink1 is required for mitochondrial function and interacts genetically with Parkin.
      ,
      • Park J.
      • et al.
      Mitochondrial dysfunction in Drosophila PINK1 mutants is complemented by Parkin.
      ,
      • Narendra D.
      • Tanaka A.
      • Suen D.F.
      • Youle R.J.
      Parkin is recruited selectively to impaired mitochondria and promotes their autophagy.
      ]. Importantly, mitochondrial dysfunction and reduced rates of mitophagy are evident in sporadic PD [
      • Keeney P.M.
      • Xie J.
      • Capaldi R.A.
      • Bennett J.P.
      Parkinson's disease brain mitochondrial complex I has oxidatively damaged subunits and is functionally impaired and misassembled.
      ,
      • Ryan B.J.
      • Hoek S.
      • Fon E.A.
      • Wade-Martins R.
      Mitochondrial dysfunction and mitophagy in Parkinson's: from familial to sporadic disease.
      ,
      • Bose A.
      • Beal M.F.
      Mitochondrial dysfunction in Parkinson's disease.
      ,

      Requejo-Aguilar, R. & Bolanos, J. P. Mitochondrial control of cell bioenergetics in Parkinson's disease. Free radical biology & medicine 100, 123-137, doi:10.1016/j.freeradbiomed.2016.04.012 (2016).

      ,
      • Luo Y.
      • Hoffer A.
      • Hoffer B.
      • Qi X.
      Mitochondria: a therapeutic target for Parkinson's disease?.
      ,
      • Hsieh C.H.
      • et al.
      Functional impairment in miro degradation and mitophagy is a shared feature in familial and sporadic Parkinson's disease.
      ,
      • Toomey C.E.
      • et al.
      Mitochondrial dysfunction is a key pathological driver of early stage Parkinson's.
      ,
      • Soutar M.P.M.
      • et al.
      Regulation of mitophagy by the NSL complex underlies genetic risk for Parkinson's disease at 16q11.2 and MAPT H1 loci.
      ], and variants in genes associated with mitochondrial function are associated with PD risk [

      Billingsley K.J., et al. Mitochondria function associated genes contribute to Parkinson’s Disease risk and later age at onset. NPJ Parkinsons Dis 2019 May 22;5:8. doi:10.1038/s41531-019-0080-x.

      ], again linking mitochondrial health and clearance processes to PD pathophysiology. Taken together, enhancing mitochondrial clearance by mitophagy is a promising disease-modifying strategy in PD.
      The PINK1-Parkin mitochondrial quality control system is a well-studied pathway in PD. In healthy cells, the kinase PINK1 is targeted to mitochondria and N-terminally translocated to the inner mitochondrial membrane (IMM). Sequential proteolysis and proteasomal degradation maintain low basal levels of PINK1 protein [
      • Yamano K.
      • Youle R.J.
      PINK1 is degraded through the N-end rule pathway.
      ]. Mitochondrial damage, typically presenting as loss of mitochondrial membrane potential (MMP), stabilizes the active PINK1 protein on the outer mitochondrial membrane (OMM) [
      • Narendra D.P.
      • et al.
      PINK1 is selectively stabilized on impaired mitochondria to activate Parkin.
      ]. Stabilized PINK1 auto-phosphorylates, homodimerizes, and then phosphorylates serine 65 (Ser65) residues of ubiquitin present at the mitochondrial surface under basal conditions [
      • Okatsu K.
      • et al.
      A dimeric PINK1-containing complex on depolarized mitochondria stimulates Parkin recruitment.
      ,
      • Okatsu K.
      • et al.
      PINK1 autophosphorylation upon membrane potential dissipation is essential for Parkin recruitment to damaged mitochondria.
      ,
      • Kondapalli C.
      • et al.
      PINK1 is activated by mitochondrial membrane potential depolarization and stimulates Parkin E3 ligase activity by phosphorylating Serine 65.
      ,
      • Sulkshane P.
      • et al.
      Inhibition of proteasome reveals basal mitochondrial ubiquitination.
      ]. The E3-ubiquitin ligase Parkin binds phospho-Ser65-ubiquitin (p-Ser65-Ub) and is phosphorylated by PINK1 on the homologous Ser65 residue of its own ubiquitin-like (UBL) domain [
      • Kondapalli C.
      • et al.
      PINK1 is activated by mitochondrial membrane potential depolarization and stimulates Parkin E3 ligase activity by phosphorylating Serine 65.
      ,
      • Kane L.A.
      • et al.
      PINK1 phosphorylates ubiquitin to activate Parkin E3 ubiquitin ligase activity.
      ,
      • Kazlauskaite A.
      • et al.
      Binding to serine 65-phosphorylated ubiquitin primes Parkin for optimal PINK1-dependent phosphorylation and activation.
      ]. These binding events and post-translational modifications release the auto-inhibitory conformation of Parkin, stabilizing an active conformation [
      • Kazlauskaite A.
      • et al.
      Binding to serine 65-phosphorylated ubiquitin primes Parkin for optimal PINK1-dependent phosphorylation and activation.
      ,
      • Gladkova C.
      • Maslen S.L.
      • Skehel J.M.
      • Komander D.
      Mechanism of Parkin activation by PINK1.
      ]. Parkin further ubiquitinates mitochondrial surface proteins [
      • Sarraf S.A.
      • et al.
      Landscape of the PARKIN-dependent ubiquitylome in response to mitochondrial depolarization.
      ], facilitating a feed-forward amplification loop of substrate ubiquitination, phosphorylation, and Parkin recruitment to tag damaged mitochondria [
      • Narendra D.
      • Tanaka A.
      • Suen D.F.
      • Youle R.J.
      Parkin is recruited selectively to impaired mitochondria and promotes their autophagy.
      ,
      • Kazlauskaite A.
      • et al.
      Binding to serine 65-phosphorylated ubiquitin primes Parkin for optimal PINK1-dependent phosphorylation and activation.
      ,
      • Wauer T.
      • Simicek M.
      • Schubert A.
      • Komander D.
      Mechanism of phospho-ubiquitin-induced PARKIN activation.
      ,
      • Ordureau A.
      • et al.
      Quantitative proteomics reveal a feedforward mechanism for mitochondrial PARKIN translocation and ubiquitin chain synthesis.
      ]. Mitochondrial surface proteins Miro and mitofusin1/2 are targeted for proteasomal degradation to modulate mitochondrial dynamics [
      • Sarraf S.A.
      • et al.
      Landscape of the PARKIN-dependent ubiquitylome in response to mitochondrial depolarization.
      ,
      • Tanaka A.
      • et al.
      Proteasome and p97 mediate mitophagy and degradation of mitofusins induced by Parkin.
      ,
      • Wang X.
      • et al.
      PINK1 and Parkin target Miro for phosphorylation and degradation to arrest mitochondrial motility.
      ], and poly-ubiquitinated OMM proteins act as receptors for autophagic adaptors and permit autophagosome formation for removal by the lysosome [
      • Lazarou M.
      • et al.
      The ubiquitin kinase PINK1 recruits autophagy receptors to induce mitophagy.
      ].
      Parkin activity is opposed by ubiquitin specific peptidase 30 (USP30), a mitochondrial deubiquitinase (DUB) which removes ubiquitin from poly-ubiquitinated proteins on mitochondria, preventing mitophagy [
      • Marcassa E.
      • et al.
      Dual role of USP30 in controlling basal pexophagy and mitophagy.
      ]. Indeed, loss or inhibition of USP30 enhances both stress-induced and basal mitophagy [
      • Marcassa E.
      • et al.
      Dual role of USP30 in controlling basal pexophagy and mitophagy.
      ,
      • Rusilowicz-Jones E.V.
      • et al.
      USP30 sets a trigger threshold for PINK1-PARKIN amplification of mitochondrial ubiquitylation.
      ,
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ,

      Rusilowicz-Jones E.V., et al. Benchmarking a highly selective USP30 inhibitor for enhancement of mitophagy and pexophagy. Life Sci Alliance 2021 Nov 29;5(2):e202101287. doi:10.26508/lsa.202101287.

      ,
      • Bingol B.
      • et al.
      The mitochondrial deubiquitinase USP30 opposes parkin-mediated mitophagy.
      ]. USP30 has lower catalytic activity against phosphorylated ubiquitin linkages and likely acts upstream of PINK1, together suggesting that USP30 activity may set the threshold for mitophagy initiation [
      • Marcassa E.
      • et al.
      Dual role of USP30 in controlling basal pexophagy and mitophagy.
      ,
      • Wauer T.
      • et al.
      Ubiquitin Ser65 phosphorylation affects ubiquitin structure, chain assembly and hydrolysis.
      ,
      • Ordureau A.
      • et al.
      Global landscape and dynamics of Parkin and USP30-dependent ubiquitylomes in iNeurons during mitophagic signaling.
      ]. Inhibition of USP30, and the consequent reduction in mitophagy initiation threshold, may help preserve mitochondrial quality by accelerating removal of damaged or deteriorating mitochondria, making it an attractive target for modulating PINK1-Parkin-dependent mitophagy in PD.
      The abundance of p-Ser65-Ub is a function of both total ubiquitination state of the mitochondria, as determined by Parkin activity, and ubiquitin phosphorylation status, as determined by PINK1 activity. As PINK1 is the only recognized Ser65-ubiquitin kinase [
      • Kane L.A.
      • et al.
      PINK1 phosphorylates ubiquitin to activate Parkin E3 ubiquitin ligase activity.
      ,
      • Schubert A.F.
      • et al.
      Structure of PINK1 in complex with its substrate ubiquitin.
      ], mitochondrial stress-induced phosphorylation of ubiquitin (p-Ser65-Ub) represents a key biomarker of mitophagy initiation. Here, we established a target-agnostic, high-content phenotypic screen (HCS) to identify small molecule enhancers of p-Ser65-Ub in a familial PD human fibroblast carrying a heterozygous PRKN R275W point mutation. We screened ∼125,000 compounds from the Eisai compound library and, following hit confirmation and counter-screening, identified compounds displaying desirable activity and drug-like properties to commence a drug discovery campaign. These compounds were further prioritized based on pharmacological profile and assessment of compound-driven functional en masse mitochondrial clearance. Biochemical screening of hits led to identification of putative inhibitors of the mitophagy negative regulator USP30, thus validating our approach. We believe that the shared mechanistic basis between HCS assay endpoint and disease phenotype creates robust predictive power for therapeutic development, enabling identification of novel small molecule modifiers of mitophagy in PD.

      2. Materials and methods

      2.1 Materials and compound synthesis

      Compounds tested in the HCS were derived from the Eisai compound library. All other chemicals and compounds were purchased from Merck unless otherwise specified. All cell culture media and supplements were purchased from Thermo Fisher Scientific unless otherwise specified. Putative Parkin activators (PPAs) were derived from patent ID: WO 2018/023029 and synthesized in-house; structures are presented in Supplementary Table 2. Details of the compounds included in the proof-of-concept study (Fig. 3) can be found in Supplementary Table 3.

      2.2 Antibodies

      For indirect immunocytochemistry (ICC), the following antibodies were used: mouse anti-PINK1 (Novus Biologicals, NBP236488, 1:1000). For direct ICC the following conjugated antibodies were used: Alexa Fluor 647-labelled anti-p-Ser65-Ub (Cell Signalling Technology [CST], 62802, 1:1500), Alexa Fluor 488-labelled anti-HSP60 (Abcam, ab128567, 1:1000), and Alexa Fluor 555 labelled-anti-LAMP1 (BD Biosciences, 555798, 1:1000). Anti-p-Ser65-Ub was conjugated by CST using their basic conjugation service. Anti-HSP60 and anti-LAMP1 were conjugated using Lightning Link Alexa 488 (Abcam, ab236553) and Alexa 555 (Abcam, ab269820) respectively, following the manufacturer's protocol, with an overnight incubation at room temperature.

      2.3 Cell culture

      Control and patient-derived fibroblasts were obtained from the NINDS Human Cell and Data Repository (https://stemcells.nindsgenetics.org). Two cell lines were primarily used: Parkin+/+ unaffected healthy control (ND36320; labelled as ‘+/+ (1)’ in the fibroblast characterisation panel in Fig. 1B) and Parkin +/R275W (ND29369; ‘+/R275W’; used for the HCS). The following additional fibroblast lines were used (Fig. 1B): Parkin +/+ (ND34769 [+/+ (2)]), Parkin +/+ (ND34770 [+/+ (3)]), Parkin N52fs/Ex3-4Del (ND29543), Parkin R42P/Ex3Del (ND30171), Parkin R42P/+ (ND31618), Parkin Ex4-7Del/Q43fs (ND40067), Parkin R275W/R275Q (ND40078), and Parkin R245K/G430D (NN0004771/NH50289). These fibroblast lines are summarized in Supplementary Table 1. All fibroblasts were sequenced for PRKN (Genewiz), verifying the above genotypes. Cell expansion to provide sufficient biomass for the HCS was performed by SAL Scientific.
      Fig 1
      Fig. 1Development of a p-Ser65-Ub high-content phenotypic assay. A, p-Ser65-Ub response following 2 h treatment with DMSO (vehicle), 10 µM FCCP or 10 µM antimycin A/5 µg/mL oligomycin A (AA/O) in PINK1−/− and PRKN−/− SH-SY5Y cells. B, p-Ser65-Ub response following 2 h treatment with DMSO (vehicle), 10 µM FCCP, 5 µM AA/O or 200 nM valinomycin in patient-derived healthy control fibroblasts [+/+ (1), (2) and (3)] and those carrying PRKN mutations (Supplementary Table 1). C, p-Ser65-Ub response following 2 h treatment with increasing concentration of AA/O (0.1 to 10 µM) in Parkin +/R275W fibroblasts. p-Ser65 ICC was performed with increasing concentration of anti-p-Ser65-Ub antibody, with 20 µM PPA-1 (2 h pre-incubation) used to determine maximal assay window. Black box and black arrow indicate the stimulus and antibody concentration, respectively, selected for the HCS. D, p-Ser65-Ub concentration-response curves of PPA-1-4 (2 h pre-incubation) under optimized HCS assay conditions (0.5 µM AA/O) in Parkin +/R275W fibroblasts. E, p-Ser65-Ub concentration-response curves of PPA-1 and -4 (2 h pre-incubation), in both cultured and directly thawed Parkin +/R275W fibroblasts under HCS assay conditions. F, p-Ser65-Ub response in Parkin +/R275W fibroblasts in the presence and absence of the Fluoromount G (FMG) anti-fade reagent, following treatment with 20 µM PPA-1 (2 h pre-incubation) in HCS assay conditions. G, Left panel, phenotype A versus phenotype B illustration scatterplot describing spread of image texture properties of the training data set obtained using PhenoLOGIC. Each coloured dot reflects a single cell following treatment with DMSO (red) or FCCP (20 µM; green) under HCS assay conditions in Parkin +/R275W fibroblasts. 6 fields and over 200 individual cells were selected per condition. Trained algorithm was used for subsequent HCS image analysis. Right panel, representative images and application of trained algorithm of p-Ser65-Ub response in Parkin +/R275W fibroblasts following treatment with DMSO, 20 µM PPA-1 and 20 µM FCCP. Cells classified as ‘DMSO-Like’ (red) and ‘FCCP-Like’ (green) phenotypes. Scale bar 200 µm. H, Scatterplot, correlation coefficient (r2) and line of best fit for p-Ser65-Ub normalized response between initial and repeat (15 h post-processing) image acquisition of the same plate. I, p-Ser65-Ub response of positive (PPA-1 and FCCP; both 20 µM) and negative controls (DMSO) in Parkin +/R275W fibroblasts under HCS assay and analysis conditions. J, Determination of the assay window using data derived in I. K, Determination of the Z' using data derived in I. For A, error bars represent s.d. (n = 3 independent experiments). *P < 0.05 and ***P < 0.001; two-way ANOVA with Bonferroni's multiple comparisons compared to isogenic control cells. For B, error bars represent s.d. (n = 3 independent experiments). For F, data are mean ± s.d. (n = 1, 8 technical replicates per condition). For I to K, data are mean ± s.d. (n = 1, 6 technical replicates per condition).
      Fibroblasts were grown in Dulbecco's Modified Eagle Medium, high glucose, with GlutaMAX (DMEM; Gibco, 61965-026), containing 10% fetal bovine serum (FBS; Biosera, FB1350-500), 1 mM sodium pyruvate (Gibco, 11360) and 100 U/mL penicillin/100 µg/mL streptomycin (Gibco, 15140). For galactose metabolism experiments, fibroblasts were grown in Dulbecco's Modified Eagle Medium with no glucose, no glutamine, no phenol red (Gibco, 12307263), containing 4.5g/L galactose (Merck, G-0625), GlutaMAX (Fisher, 11574466), 10% fetal bovine serum (dialysed, US origin, One Shot™, Fisher, A3382001), 1 mM sodium pyruvate and 100 U/mL penicillin/100 µg/mL streptomycin.
      Human neuroblastoma control SH-SY5Y cell line (Abcam, ab275475), Parkin knockout (PRKN −/−) SH-SY5Y cells (Abcam, ab280101) and PINK1 knockout (PINK1 −/−) SH-SY5Y cells (Abcam, ab280876) were cultured in Dulbecco's Modified Eagle Medium/F-12 (1:1) with GlutaMAX (DMEM/F-12; Gibco 31331-028), supplemented with 10% FBS (Biosera, FB1350-500) and 100 U/mL penicillin/100 µg/mL streptomycin.

      2.4 HCS assay: p-Ser65-Ub direct ICC

      p-Ser65-Ub was assessed using direct ICC. Freshly-thawed patient-derived fibroblasts were seeded at 3,000 cells per well in 25 µL pre-warmed complete cell culture medium in CellCarrier Ultra 384-well plates (PerkinElmer, 6057300) using a Multidrop Combi Reagent Dispenser (Thermo Scientific), allowed to adhere at room temperature for 1 h, then incubated overnight at 37 °C and under 5% CO2. Medium was replaced with 25 µL of pre-warmed complete cell culture medium containing compounds using the CyBio Cybi Well Vario (Analytik Jena) and incubated for 2 h at 37 °C and under 5% CO2. A further 25 µL of pre-warmed complete cell culture medium containing mitochondrial toxins at twice the final assay concentration (FAC; FAC of 0.5 µM antimycin A, 5 µg/mL oligomycin A) was added using the CyBio Cybi Well Vario and incubated for a further 2 h. Any deviations from these conditions are defined within figure legends. In AA/O treatments, oligomycin A was always used at a FAC of 5 µg/mL.
      Medium was removed and cells were fixed in ice-cold acetone: methanol (A:M; 1:1) for 30 seconds, added using the CyBio Cybi Well Vario. A:M was removed and 25 µL of ice-cold D-PBS (without Ca2+/Mg2+) was added using the Multidrop Combi Reagent Dispenser, then incubated on ice for 30 min. D-PBS was replaced with ice-cold blocking solution (3% BSA [Sigma, A3803] in D-PBS, without Ca2+/Mg2+) using a Dragonfly Discovery liquid dispenser (SPT Labtech) and incubated for 1 h on ice. Blocking solution was replaced with ice-cold blocking solution (3% BSA in D-PBS) containing conjugated antibodies (Alexa 647 anti-p-Ser65-Ub [CST, 62802, 1:1500], Alexa 488 anti-HSP60 [Abcam, ab128567, 1:1000], and Alexa 555 anti-LAMP1 [BD Biosciences, 555798, 1:1000]) using a Dragonfly Discovery liquid dispenser and incubated at 4 °C overnight in the dark with gentle agitation.
      Cells were washed twice in room temperature D-PBS containing 0.1% Tween 20 (D-PBS-T) using the Multidrop Combi Reagent Dispenser, then 25 µL of Fluoromount G (SouthernBiotech, 0100-01): D-PBS (1:1) with Hoechst (2 µg/mL) was added per well using the Dragonfly Discovery liquid dispenser. Plates were sealed using opaque foil seals (Agilent, 24214-001) and incubated for at least 15 min at room temperature. Images were acquired using the Opera Phenix (PerkinElmer) using the 20x water objective (NA 1.0, WD 1.7 mm, field of view approximately 646 μm x 646 μm) and 4 fields imaged per well.

      2.5 p-Ser65-Ub direct ICC in SH-SY5Y cells

      p-Ser65-Ub was assessed in SH-SY5Y cells using direct ICC as detailed for the HCS assay but with minor modifications. SH-SY5Y cells were seeded at 12,500 cells per well in pre-warmed complete cell culture medium in CellCarrier Ultra 384-well plates using the Dragonfly Discovery liquid dispenser and incubated overnight at 37 °C and under 5% CO2. Medium was replaced with 25 µL of pre-warmed complete cell culture medium containing treatments using the CyBio Cybi Well Vario and incubated for 2 h at 37 °C and under 5% CO2. In AA/O treatments, oligomycin A was always used at a final assay concentration of 5 µg/mL.
      Medium was removed and cells were fixed in ice-cold A:M (1:1) for 30 seconds, added using the CyBio Cybi Well Vario. A:M was removed and 25 µL of ice-cold D-PBS (without Ca2+/Mg2+) was added using the Cybio Cybi Well Vario, then incubated on ice for 30 min. D-PBS was replaced with ice-cold blocking solution (3% BSA in D-PBS) using the Dragonfly Discovery liquid dispenser and incubated for 1 h on ice. Blocking solution was replaced with ice-cold blocking solution (3% BSA in D-PBS) containing all conjugated antibodies (Alexa 647 anti-p-Ser65-Ub [CST, 62802, 1:1500], Alexa 488 anti-HSP60 [Abcam, ab128567, 1:1000], and Alexa 555 anti-LAMP1 [BD Biosciences, 555798, 1:1000]) using the Dragonfly Discovery liquid dispenser and incubated at 4 °C overnight in the dark with gentle agitation.
      Cells were washed twice in room temperature D-PBS with 0.1% Tween 20 (D-PBS-T) using the Cybio Cybi Well Vario, then 25 µL of Fluoromount G: D-PBS (1:1) with Hoechst (2 µg/mL) was added per well using the Dragonfly Discovery liquid dispenser. Plates were sealed using opaque foil seals and incubated for at least 15 min at room temperature before image acquisition. Images were acquired using the Opera Phenix using the 20x water objective (NA 1.0, WD 1.7 mm, field of view approximately 646 μm x 646 μm) and 4 fields imaged per well.

      2.6 Mitochondrial membrane potential assay: TMRM

      Mitochondrial membrane potential was measured using tetramethyl rhodamine methyl ester (TMRM) in redistribution mode [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ]. Freshly-thawed patient-derived fibroblasts were seeded at 3,000 cells per well in CellCarrier Ultra 384-well plates, allowed to adhere at room temperature for 1 h, then incubated overnight at 37 °C and under 5% CO2. Medium was replaced with 25 µL of pre-warmed complete cell culture medium containing compounds using CyBio Cybi Well Vario and incubated for 2 h. A further 25 µL of pre-warmed complete cell culture medium containing mitochondrial toxins (FAC of 0.5 µM antimycin A, 5 µg/mL oligomycin A) in TMRM staining solution (FAC of 50 nM TMRM, 1 µg/mL Hoechst) was added using CyBio Cybi Well Vario and incubated for 2 h at 37 °C to allow equilibration of dye. In AA/O treatments, oligomycin A was always used at a final assay concentration of 5 µg/mL. Images were acquired using the Opera Phenix, 20x water objective (NA 1.0, WD 1.7 mm, field of view approximately 646 μm x 646 μm), 2 fields imaged per well, with temperature and CO2 controls enabled.

      2.7 Functional mitophagy assay: mito-eGFP clearance

      Fibroblasts were transfected using custom mRNA (provided by Trilink Biotechnologies). Mitochondrial targeted-enhanced Green Fluorescent Protein (Mito-eGFP) mRNA was designed by appending a N-terminal mitochondrial localisation sequence derived from COX8a (accession number: NP_004065.1) to eGFP. 25 µg mRNA was transfected into 2.5  ×  106 freshly-thawed fibroblasts using Cell Line Nucleofector Kit V (Lonza) and AMAXA program X-001 following the manufacturer's instructions. Cells were seeded at 5,000 cells per well in CellCarrier Ultra 384-well plates, allowed to adhere at room temperature for 1 h, then incubated overnight at 37 °C and under 5% CO2.
      Compounds were added in a full medium change at 20 h following transfection and cell seeding. Medium was replaced with 25 µL of pre-warmed complete cell culture medium containing compounds using the CyBio Cybi Well Vario and incubated for a further 2 h. A further 25 µL of pre-warmed complete cell culture medium containing mitochondrial toxins at twice the FAC (5-point concentration range of antimycin A [FAC of 0.04-5 µM] with oligomycin A [constant FAC of 5 µg/mL] and Hoechst [FAC of 1 µg/mL]) was added using CyBio Cybi Well Vario and incubated for 28 h. eGFP signal was acquired using the Opera Phenix, 20x water objective (NA 1.0, WD 1.7 mm, field of view approximately 646 μm x 646 μm), 3 fields imaged per well, with temperature and CO2 controls enabled.
      Following imaging of eGFP, a further 5 µL of pre-warmed complete cell culture medium containing 6x TMRM staining solution (300 nM TMRM; FAC of 50 nM) was added using the Dragonfly Discovery liquid dispenser and incubated for 1 h at 37 °C to allow equilibration of dye. Images were acquired using the Opera Phenix, 20x water objective (NA 1.0, WD 1.7 mm, field of view approximately 646 μm x 646 μm), 2 fields per well, with temperature and CO2 controls enabled. Following live-cell imaging, cells were fixed and immuno-stained for PINK1, following the indirect ICC method below.

      2.8 Indirect ICC

      For fibroblasts, medium was removed, and cells were fixed in ice-cold A:M (1:1) for 30 seconds, added using CyBio Cybi Well Vario. A:M was removed and 25 µL of ice-cold D-PBS (without Ca2+/Mg2+) was added using Multidrop Combi Reagent Dispenser, then incubated on ice for 30 min. D-PBS was replaced with ice-cold blocking solution (3% BSA in D-PBS) using the Dragonfly Discovery liquid dispenser and incubated for 1 h on ice. Blocking solution was replaced with ice-cold blocking solution (3% BSA in D-PBS) containing appropriate antibodies using the Dragonfly Discovery liquid dispenser and incubated at 4 °C overnight with gentle agitation.
      Cells were washed three times in room temperature TBS with 0.1% Tween 20 (TBS-T) using Multidrop Combi Reagent Dispenser, then room temperature blocking solution (3% BSA in D-PBS) containing secondary antibody (IgG2b Cross-Adsorbed Goat anti-Mouse, Alexa Fluor™ 555 [Invitrogen, 10412832], 1:2000) was added using Dragonfly Discovery liquid dispenser and incubated for 1 h at room temperature. Cells were again washed three times in room temperature TBS-T using the Multidrop Combi Reagent Dispenser, then 25 µL of Fluoromount G: TBS (1:1) with Hoechst (2 µg/mL) was added per well using the Dragonfly Discovery liquid dispenser. Plates were sealed using opaque foil seals and incubated for at least 15 min at room temperature before image acquisition. Images were acquired using the Opera Phenix, 40x water objective (NA 1.1, WD 0.62 mm, field of view approximately 323 μm x 323 μm), 8 fields imaged per well.

      2.9 USP30-Ub Rho110 activity assay

      USP30-UbRho10 was performed as previously described [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ]. Compounds were dispensed into black, clear-bottom, 384-well plates (Greiner, 781096) using the ECHO 550 (Labcyte) liquid handler. 2 × FAC His-tagged recombinant human USP30 protein (rhUSP30; amino acids 57–517 of the full-length protein, and a C-terminal 6-His tag, Sf 21 (baculovirus)-derived; 10 nM, FAC of 5 nM [R&D Systems, E-582-050]) was prepared in USP30 activity assay buffer (50 mM Tris base pH 7.5, 100 mM NaCl, 0.1 mg/mL BSA [Sigma, A7030], 0.05% Tween 20, 1 mM DTT) and 15 µL dispensed into compound-containing assay plate using the Dragonfly Discovery liquid dispenser and incubated for 30 min at room temperature. Following incubation, 15 µL of 2x concentrated ubiquitin-rhodamine 110 (Ub-Rho110; 200 nM, FAC of 100 nM [R&D Systems, U-555-050]) in USP30 activity assay buffer was dispensed into the compound-rhUSP30 containing plate using the Dragonfly Discovery liquid dispenser and fluorescence read on the FLIPR TETRA plate reader (Molecular Devices). Fluorescence intensity was recorded over 1 h and normalized to positive and negative controls (USP30Inh-1 [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ], 10 µM and DMSO respectively).

      2.10 Deubiquitinase selectivity profiling

      Selectivity profiling against forty one deubiquitinase (DUB) enzymes was performed at Ubiquigent (Dundee, U.K.) using the DUBprofiler™ platform and Ub-Rho110-glycine substrate based-assay.

      2.11 Data analysis and statistics

      Data are presented as mean ± standard deviation (SD). Normalization of the data allowed for control of inter-assay variability. Assay quality was determined after calculation of Z′, using the equation: Z=13(σp+σn)μpμn, with standard deviation (σ) and means (μ) of the positive (p) and negative (n) controls [
      • Zhang J.H.
      • Chung T.D.
      • Oldenburg K.R.
      A simple statistical parameter for use in evaluation and validation of high throughput screening assays.
      ].
      For the HCS p-Ser65-Ub ICC assay, image analysis was performed using the Perkin Elmer Harmony PhenoLOGIC machine learning plug-in, which was used to score cells as either p-Ser65-Ub positive or p-Ser65-negative, based on the image texture phenotype of positive and negative controls. PhenoLOGIC plug-in enables the recognition of phenotypes within cell populations using a machine learning approach that combines the most meaningful parameters to achieve accurate classification of cell phenotypes at a single cell level. Data are presented as percentage positive cells normalized to positive and negative controls (FCCP, 20 µM FAC, and DMSO respectively). Plates with Z′ ≥ 0.8 and secondary QC ≥ 20% were considered to pass QC. For the TMRM assay, Z′≥ 0.8 was considered a pass. For all other high-content imaging, including hit compound pharmacology in p-Ser65-Ub assay, fluorescence intensity of signal was calculated per nuclei number and normalized to positive and negative controls.
      Data handling, statistical analysis, and data visualisation were performed using Graphpad Prism 9 and Tibco Spotfire Analyst 10.3.3. Curve fitting was performed using the IDBS XLfit (5.4.0.8) add-in for Microsoft Excel. Statistical tests are indicated in figure legends. Statistical significance was assessed as being P < 0.05.
      For the mito-eGFP functional mitophagy assay, data were normalized to positive and negative controls (200 µM valinomycin and DMSO respectively). Area under the curve (AUC) over the 5-point AA/O concentration range was calculated for both test compound and 6x DMSO technical replicates per plate using the IDBS XLfit (5.4.0.8) add-in for Microsoft Excel and presented as fold change relative to average DMSO AUC (compound AUC/DMSO AUC).
      Chemical profile of the selected hit compounds was generated by ChemAxon v16.10.24.

      3. Results

      3.1 Development of a p-Ser65-Ub high-content phenotypic assay

      To confirm our methodology and the contribution of PINK1-Parkin signalling to the accumulation of p-Ser65-Ub, we first quantified mitochondrial stress-induced phosphorylation of ubiquitin in control, Parkin −/− and PINK1 −/− SH-SY5Y cells via immunocytochemistry (ICC; Fig. 1A). A marked increase in p-Ser65-Ub levels – indicative of PINK1-Parkin pathway activation – was observed in isogenic control cells following incubation with FCCP or antimycin A/oligomycin A (AA/O), which was significantly reduced in Parkin −/− cells, and abolished in PINK1 −/− cells (Fig. 1A), suggesting p-Ser65-Ub is a reliable and specific reporter of PINK1-Parkin function and is both detectable and quantifiable using ICC.
      Previously, we identified patient-derived familial PD fibroblasts which demonstrate mitophagy deficits and decreased p-Ser65-Ub accumulation, with a clear genotype-phenotype relationship determined by allelic number of PRKN mutations [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ]. Furthering these observations, we profiled p-Ser65-Ub responses in an expanded panel of patient-derived Parkin mutant fibroblasts (Fig. 1B). We found significant deficits in p-Ser65-Ub accumulation associated with severity of PRKN genotype across a range of mitophagy-inducing stimuli, with compound heterozygous mutants being most affected (Fig. 1B). These data suggest a strong relationship between PRKN status and stress-induced p-Ser65-Ub accumulation and, together with evidence of mitochondrial dysfunction in sporadic PD [
      • Keeney P.M.
      • Xie J.
      • Capaldi R.A.
      • Bennett J.P.
      Parkinson's disease brain mitochondrial complex I has oxidatively damaged subunits and is functionally impaired and misassembled.
      ,
      • Ryan B.J.
      • Hoek S.
      • Fon E.A.
      • Wade-Martins R.
      Mitochondrial dysfunction and mitophagy in Parkinson's: from familial to sporadic disease.
      ,
      • Bose A.
      • Beal M.F.
      Mitochondrial dysfunction in Parkinson's disease.
      ,

      Requejo-Aguilar, R. & Bolanos, J. P. Mitochondrial control of cell bioenergetics in Parkinson's disease. Free radical biology & medicine 100, 123-137, doi:10.1016/j.freeradbiomed.2016.04.012 (2016).

      ,
      • Luo Y.
      • Hoffer A.
      • Hoffer B.
      • Qi X.
      Mitochondria: a therapeutic target for Parkinson's disease?.
      ,
      • Hsieh C.H.
      • et al.
      Functional impairment in miro degradation and mitophagy is a shared feature in familial and sporadic Parkinson's disease.
      ,
      • Toomey C.E.
      • et al.
      Mitochondrial dysfunction is a key pathological driver of early stage Parkinson's.
      ,
      • Soutar M.P.M.
      • et al.
      Regulation of mitophagy by the NSL complex underlies genetic risk for Parkinson's disease at 16q11.2 and MAPT H1 loci.
      ], provide rationale for the development of a p-Ser65-Ub endpoint assay to identify molecules augmenting PINK1-Parkin signalling to normalize mitochondrial deficits.
      We next developed a high-content phenotypic screening assay to classify compounds capable of enhancing p-Ser65-Ub accumulation. Parkin +/R275W fibroblasts have a reduced p-Ser65-Ub response compared to common variant controls (Fig. 1B) and, importantly, we have previously shown enhanced p-Ser65-Ub accumulation in these cells following USP30 inhibition, suggesting that p-Ser65-Ub can be modulated via pharmacological intervention of the PINK1-Parkin pathway [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ]. As such, a high-content imaging assay for p-Ser65-Ub was established in Parkin +/R275W fibroblasts, with the aim of analyzing small molecule-mediated normalization of mitophagy deficits in a familial PD cellular background. In addition to p-Ser65-Ub, HSP60 and LAMP1 were included as markers of mitochondrial and lysosomal compartments, respectively.
      Following automation and miniaturization of the p-Ser65-Ub ICC assay to a 384-well plate format, we optimized assay conditions. We selected AA/O treatment over 2 h incubation as the acute mitophagy-inducing stimulus for our assay. We titrated AA and antibody concentration to determine the largest assay window in the presence of putative Parkin activator-1 (PPA-1; derived from patent WO 2018/023029) (Fig. 1C). A three-fold change in p-Ser65-Ub with PPA-1 was measured with 0.5 µM AA (in the presence of constant 5 µg/mL oligomycin A) using the fluorophore-conjugated p-Ser65-Ub (E2J6T)-Alexa647 at 1:1500 dilution, resulting in a large assay window (Fig. 1C). Higher antibody concentrations failed to show a p-Ser65-Ub change, and lower concentrations (1:2000 dilution) produced greater signal variability (Fig. 1C). At lower AA/O concentrations a small p-Ser65-Ub response was observed, and high AA/O concentrations revealed a saturated p-Ser65-Ub response that could not be enhanced by PPA-1, each limiting the assay window (Fig. 1C). To confirm our assay was fit-for-purpose, putative Parkin activators (PPA-1-4) were tested under optimized assay conditions, obtaining a concentration-response for all four compounds (Fig. 1D). PPA-1, which produced the largest p-Ser65-Ub response, was selected as a mechanistically relevant positive control for the HCS.
      To improve efficiency and consistency throughout the screen, we identified that direct thaw of cryopreserved cells produced equivalent pharmacological profiles for PPA-1 and PPA-4 compared to cells maintained in culture (Fig. 1E). As such, cells were batch prepared to generate over 1 × 109 cells, enough to complete primary and counter-screening, allowing us to ensure cells were passage-matched throughout the HCS and to minimize impact of variation in culture conditions. The addition of the anti-fade reagent Fluoromount G (FMG; diluted 1:1 in D-PBS) following antibody incubation resulted in improved p-Ser65-Ub signal intensity and preservation (Fig. 1F) and was used routinely.
      An image analysis method was developed using the Harmony PhenoLOGIC machine learning plug-in (PerkinElmer) to score cells as either p-Ser65-Ub positive or p-Ser65-Ub negative, based on the image texture phenotype of the positive (FCCP) and negative (DMSO) controls (Fig. 1G). FCCP was selected as the positive control given the reliable enhancement of p-Ser65-Ub in these cells (Fig. 1B). In the presence of FMG and using the PhenoLOGIC analysis, assay readout was stable across 15 h at room temperature (the longest duration plates were waiting to be imaged in a single run; Fig. 1H).
      Finally, we validated FCCP and PPA-1, our two positive controls in the assay (Fig. 1I-K). FCCP produced a large p-Ser65-Ub signal (Fig. 1I) and assay window (Fig. 1J), and a Z-Prime (Z'; a statistical measure of assay robustness [
      • Zhang J.H.
      • Chung T.D.
      • Oldenburg K.R.
      A simple statistical parameter for use in evaluation and validation of high throughput screening assays.
      ]) between 0.8-0.9, indicating a high-quality assay (Fig. 1K). PPA-1 produced a lower p-Ser65-Ub signal, assay window, and Z' (Fig. 1I-K) but was more mechanistically relevant (a putative mitophagy enhancer rather than a mitochondrial toxin). On this basis, we selected two assay quality control (QC) measures: primary QC was the plate Z', using the DMSO negative control and FCCP as the primary positive control; secondary QC was the normalized p-Ser65-Ub response to PPA-1, used to confirm run-to-run robustness of assay sensitivity. The p-Ser65-Ub response was normalized based on the negative (DMSO; 0%) and positive (FCCP; 100%) response (percentage positive cells; Fig. 1G). Plates with a Z' above 0.8 and a secondary QC above 20% normalized response satisfied our screening criteria and were used for further analysis. Based on the response of PPA-1 and the distribution of negative control responses, we selected a ‘hit’ threshold of 15% normalized p-Ser65-Ub response. Thorough optimization of the p-Ser65-Ub ICC assay protocol allowed us to establish a robust, quality-controlled, and reproducible HCS assay.

      3.2 Establishment of a TMRM counter-screen to exclude compounds that affect mitochondrial membrane potential

      Mitochondrial function is critically linked to mitochondrial polarization state. Mitochondrial stress or damage typically presents in vitro as loss of MMP triggering PINK1-Parkin mitophagy and the accumulation of p-Ser65-Ub [
      • Narendra D.
      • Tanaka A.
      • Suen D.F.
      • Youle R.J.
      Parkin is recruited selectively to impaired mitochondria and promotes their autophagy.
      ,
      • Jin S.M.
      • et al.
      Mitochondrial membrane potential regulates PINK1 import and proteolytic destabilization by PARL.
      ]. To understand the relationship between MMP depolarization and p-Ser65-Ub accumulation in our HCS assay, we examined MMP by tetramethyl-rhodamine methyl ester (TMRM) fluorescence in a well-validated and established imaging assay [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ]. We demonstrated an inverse correlation between TMRM intensity and p-Ser65-Ub accumulation upon treatment with AA/O (Fig. 2A-B). Following 2 h incubation with AA/O (analogous to HCS conditions), a p-Ser65-Ub response was only observed at concentrations of AA/O which caused a >75% reduction in TMRM fluorescence (Fig. 2B). Our HCS-stimulus condition (0.5 µM AA/5 µg/mL oligomycin A) reduced TMRM fluorescence by approximately 60% and did not induce a substantial p-Ser65-Ub percentage positive cell response (Fig 2A-B).
      Fig 2
      Fig. 2Establishment of a TMRM counter-screen to exclude compounds that affect mitochondrial membrane potential. A, Representative images of p-Ser65-Ub and TMRM following 2 h antimycin A (with constant oligomycin A; 5 µg/mL) concentration-response in Parkin +/R275W fibroblasts. Pink box indicates the antimycin A concentration [AA] selected for HCS. Scale bar 500 µm. B, Quantification of TMRM fluorescence intensity and percentage of p-Ser65-Ub positive cells as in A. Vertical dotted line indicates the antimycin A concentration [AA] selected for HCS. C-E, Scatter plots of p-Ser65-Ub normalized response of test compound (blue dots) versus TMRM normalized response following: C, 2 h treatment with compound in glucose-containing DMEM; D, Standard assay conditions of 2 h pre-treatment with compounds followed by 2 h treatment with 0.5 µM AA/O; and E, 2 h treatment with compound in galactose-containing DMEM. In C, D, and E, number of tested compounds (dots) is 344. Black dotted lines mark the 15% p-Ser65-Ub normalized response hit threshold. F, Hit rate (percentage of the 344 test compounds which pass both the p-Ser65-Ub threshold (> 15%) and the indicated TMRM threshold) is reported applying different TMRM cut-offs (10, 40 and 75% TMRM normalized response).
      Compounds which inherently depolarize MMP or have a cumulative depolarizing effect when combined with AA/O would increase p-Ser65-Ub and were designated as ‘hits’ in our HCS assay, despite acting via a nonspecific mechanism. To identify and exclude these undesirable compounds, we developed a TMRM protocol for use as a counter-screening assay (Fig. 2C-F). A representative set of compounds were screened using the p-Ser65-Ub HCS assay, and the TMRM assay was performed in parallel after either 2 h treatment with compounds (Fig. 2C); 2 h treatment with compounds followed by 2 h treatment with AA/O (Fig. 2D), matching the p-Ser65-Ub HCS conditions; or 2 h treatment with compounds in galactose-substituted media (Fig. 2E), following overnight incubation with galactose, to metabolically force the cells into oxidative phosphorylation instead of glycolysis and make them more sensitive to mitochondrial insult [
      • Marroquin L.D.
      • Hynes J.
      • Dykens J.A.
      • Jamieson J.D.
      • Will Y.
      Circumventing the Crabtree effect: replacing media glucose with galactose increases susceptibility of HepG2 cells to mitochondrial toxicants.
      ]. Of the 344 compounds tested, 32 compounds hit in the p-Ser65-Ub HCS assay (hit rate = ∼10%). When tested in glucose-containing medium without AA/O, most hit compounds had limited effect on TMRM fluorescence (Fig. 2C). However, in the presence of AA/O or in galactose-substituted media, more compounds reduced TMRM fluorescence, indicating an effect on MMP (Fig. 2D-E).
      As shown in Figure 2B, a reduction in TMRM of approximately 75% is necessary for a p-Ser65-Ub response in our HCS fibroblast model. However, this may not be a sufficiently harsh threshold for compound selection as MMP depolarization is a major confounding factor for mitophagy activation and must be avoided. To select a suitable threshold for the TMRM counter-screening assay, we compared hit rates in the three TMRM assay conditions using different exclusion thresholds (Fig. 2F). A relaxed TMRM threshold (40% or 75%) only slightly reduced the 10% hit rate obtained from the p-Ser65-Ub assay alone (Fig. 2F). Using a 10% TMRM cut-off, the hit rate was 5% (Fig. 2F) after 2 h incubation in glucose-containing medium. In contrast, using the same threshold in the TMRM assay in both galactose medium and under HCS conditions (with AA/O) we obtained a 1% final hit rate (Fig. 2F). We therefore selected this stringent 10% TMRM threshold and glucose medium containing AA/O as the finalized assay conditions, as this aligned the TMRM counter-screen with the primary p-Ser65-Ub HCS, permitted precise correlation between endpoints, and produced an estimated final hit rate of 1%, an acceptable outcome from a HCS [
      • Moffat J.G.
      • Vincent F.
      • Lee J.A.
      • Eder J.
      • Prunotto M.
      Opportunities and challenges in phenotypic drug discovery: an industry perspective.
      ].

      3.3 Known mitophagy modulators were active in the optimized HCS assay

      To validate our screening system, we used our optimized HCS (Fig. 1) and TMRM (Fig. 2) assays to explore the effects of published compounds proposed to affect aspects of mitochondrial and/or lysosomal biology. Forty five compounds were tested (Fig. 3A), from broad mechanistic classes including proton ionophores, kinase modulators, and proposed mitophagy activators such as PPAs and USP30 inhibitors [
      • Clark E.H.
      • Vázquez de la Torre A.
      • Hoshikawa T.
      • Briston T.
      Targeting mitophagy in Parkinson's disease.
      ]. Compounds were tested at low (1:1666 dilution), mid (1:500), and high (1:166) concentration, as determined by the starting concentration of each compound (Supplementary Table 3), and p-Ser65-Ub and TMRM endpoints measured and compared (Fig. 3B-D).
      Fig 3
      Fig. 3Known mitophagy modulators were active in the optimized HCS assay. A, Broad mechanistic class (mechanism-of-action) of the forty four literature identified compounds (plus vehicle, DMSO) employed in the proof-of-concept study. B-D, Scatter plots of p-Ser65-Ub and TMRM normalized response using HCS primary and TMRM secondary screening protocols for compounds reported in A tested at low (B), mid (C) and high (D) concentration. E, p-Ser65-Ub and TMRM normalized responses following treatment with low to high concentrations of USP30 inhibitors (USP30Inh-1, -2 and -3), derived from B-D. F, p-Ser65-Ub and TMRM normalized responses following treatment with low to high concentrations of mitophagy enhancers, derived from B-D. G, p-Ser65-Ub and TMRM normalized responses following treatment with low to high concentrations of mitochondrial toxins, derived from B-D. For E-G, height of bars indicates p-Ser65-Ub normalized response and color gradient indicates mitochondrial polarization state, as determined by TMRM intensity (polarized [blue] – depolarized [pink]). For B-D, data are mean of n = 3 independent experiments. E-G, data are mean ± s.d. (n = 3 independent experiments).
      At all concentrations, the protonophores FCCP, CCCP, and BAM15 (6-60 µM) induced strong p-Ser65-Ub responses and a high degree of MMP depolarization, whereas limited effects on either MMP or p-Ser65-Ub were observed for other compounds at low concentration (Fig. 3B). At increasing (mid and high) concentrations, several compounds demonstrated concentration-dependent effects on MMP and/or p-Ser65-Ub (Fig. 3C and D). The PPAs behaved as expected based on previous observations (Fig. 1), with PPA-1 inducing the strongest p-Ser65-Ub response, although this compound modestly impacted MMP at high concentration (60 µM; Fig. 3D). USP30 inhibitors (USP30Inh-1, 2 and 3) strongly induced p-Ser65-Ub with limited effect on MMP, as observed previously under FCCP-mediated mitochondrial stress [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ] (Fig 3E).
      Several compounds identified in recent studies as enhancers of PINK1-Parkin mitophagy also demonstrate accumulation of p-Ser65-Ub in our assay system, including T-271 (originally identified via a Mito-SRAI screen [
      • Katayama H.
      • et al.
      Visualizing and modulating mitophagy for therapeutic studies of neurodegeneration.
      ]), the ROCK2 inhibitor SR3677 (identified from a eGFP-Parkin translocation screen [
      • Moskal N.
      • et al.
      ROCK inhibitors upregulate the neuroprotective Parkin-mediated mitophagy pathway.
      ]) and the Fbxo7-FP domain inhibitor, BC1464 (identified from a virtual screen, as a modulator of Fbxo7-PINK1 interactions [
      • Liu Y.
      • et al.
      Chemical inhibition of FBXO7 reduces inflammation and confers neuroprotection by stabilizing the mitochondrial kinase PINK1.
      ]) (Fig. 3F). BC1464 only had a modest effect in our assay, however the treatment period employed in our assay system is much shorter than that used in the original paper (2 h versus 16 h compound incubation) (Fig. 3F). Furthermore, T0466, which induced Parkin translocation and enhanced Parkin-dependent degradation of luciferase-tagged MFN-1 [
      • Shiba-Fukushima K.
      • et al.
      A cell-based high-throughput screening identified two compounds that enhance PINK1-Parkin signaling.
      ], increased p-Ser65-Ub at 6 µM and 20 µM, while inducing mitochondrial depolarization at high concentration (Fig. 3F). We observed no p-Ser65-Ub response using kinetin, a proposed PINK1 activator, however previous experiments with this compound required 24 h incubation and a technology for cell delivery [
      • Osgerby L.
      • et al.
      Kinetin riboside and its ProTides activate the Parkinson's disease associated PTEN-induced putative kinase 1 (PINK1) independent of mitochondrial depolarization.
      ] (Fig. 3F). MWP00839 at high concentration (60 µM), but not SPB08007, compounds identified using Mito-Timer as enhancers of basal mitophagy [
      • Cerqueira F.M.
      • et al.
      MitoTimer-based high-content screen identifies two chemically-related benzothiophene derivatives that enhance basal mitophagy.
      ], promoted accumulation of p-Ser65-Ub, with limited effect on MMP (Fig. 3F).
      Importantly, our p-Ser65-Ub HCS assay did not identify mitophagy-enhancing compounds with mechanism-of-action independent of the PINK1-Parkin pathway it reports on, including p62-mediated mitophagy inducer (PMI), which enhances mitophagy via p62, downstream of the PINK1-Parkin axis [
      • East D.A.
      • et al.
      PMI: a ΔΨm independent pharmacological regulator of mitophagy.
      ], and the iron chelator deferiprone [
      • Allen G.F.
      • Toth R.
      • James J.
      • Ganley I.G.
      Loss of iron triggers PINK1/Parkin-independent mitophagy.
      ] (Fig. 3F). The c-ABL inhibitor nilotinib [
      • Karuppagounder S.S.
      • et al.
      The c-Abl inhibitor, nilotinib, protects dopaminergic neurons in a preclinical animal model of Parkinson's disease.
      ] failed to induce p-Ser65-Ub and demonstrated MMP depolarization at higher concentration (Fig. 3F). Interestingly, Miro1 reducer Compound 3 [
      • Hsieh C.H.
      • et al.
      Miro1 marks Parkinson's disease subset and Miro1 reducer rescues neuron loss in Parkinson's models.
      ] promoted p-Ser65-Ub accumulation at low-mid concentration in our system (Fig. 3F) but demonstrated a loss of MMP at high concentration. As predicted, mitochondrial toxins increased p-Ser65-Ub, with a heavy dependence on MMP depolarization (Fig. 3G). Together, these data established our HCS and TMRM counter-screen as fit-for-purpose and demonstrated that we could confidently use these assays to identify specific modulators of PINK1-Parkin biology.

      3.4 HCS and counter-screening identified MMP-independent positive modulators of p-Ser65-Ub

      To identify novel positive modulators of p-Ser65-Ub within the Eisai compound library, we used the finalized HCS conditions (Fig. 4A) to screen approximately 125,000 compounds at a single concentration (∼5 µg/mL, with molar concentrations ranging from 20 to 30 µM; Fig. 4B). Cells were pre-treated for 2 h with either test compound, negative (DMSO), or positive (FCCP or PPA-1) control, then AA/O (final assay concentration [FAC] of 0.5 µM AA and 5 µg/mL oligomycin A) was added as a mild mitophagy-inducing stimulus. Based on the 15% p-Ser65-Ub threshold and assay QC criteria previously established (Fig. 1), we identified approximately 15,000 hits from our primary screen, a hit rate of 12.3% (Fig. 4B-C).
      Fig 4
      Fig. 4HCS and counter-screening identified positive modulators of p-Ser65-Ub. A, Representative images of Parkin +/R275W fibroblast cells treated with either vehicle (DMSO), 20 µM FCCP, or 20 µM PPA-1 under HCS assay conditions. p-Ser65-Ub (red), HSP60 (green), LAMP1 (yellow) and Hoechst nuclear stain (blue). Scale bar 200 µm. Bottom panels are higher magnification images of the areas in the white boxes. B, Primary HCS. Normalized p-Ser65-Ub response after treatment with ∼125,000 compounds, each run in singlicate at approximately 20-30 µM, using HCS protocol. Distribution of negative control (DMSO, dark blue), positive control (20 µM FCCP, purple), secondary positive control (20 µM PPA-1, pink), and test compounds (light blue). All data presented passed QC, based on selected criteria (Z’ > 0.8 and secondary QC > 20% response). Dashed line indicates the selected hit threshold of > 15% p-Ser65-Ub response; test compounds with p-Ser65-Ub response above this threshold were selected as hits. C, Screening and counter-screening workflow. D, Hit compound counter-screening. Scatter plot of p-Ser65-Ub and TMRM normalized response using HCS primary and TMRM secondary screening protocols for ∼15,000 compounds identified in primary HCS (20 µM; singlicate; 2 h pre-incubation). Compounds were excluded based on mitochondrial membrane depolarization (± 10% change in TMRM response), failure to repeat (p-Ser65-Ub response < 15%), response in unstimulated conditions (p-Ser65-Ub response > 15% in absence of a mitochondrial stressor), and overt effects on Hoechst fluorescence. E, Representative images for the counter-screening assays. F, Selected filtering criteria for counter-screening assays and number of compounds that passed, derived from data in D.
      Primary hits were then subjected to counter-screening assays measuring four parameters (Fig. 4C-F). First, we reconfirmed p-Ser65-Ub responses using compound stocks standardized to 20 µM (p-Ser65-Ub n=2). Approximately half (7,651) of the ∼15,000 primary hit compounds were confirmed (Fig. 4D and F). We then used our optimized TMRM assay (Fig. 2), excluding both depolarizing and hyperpolarizing test compounds (Fig. 4D and E). As normalized TMRM response partially correlates with normalized p-Ser65-Ub response (r2 = 0.416; derived from data in Fig. 4D), exclusion of compounds with substantial depolarizing effect resulted in disproportionate removal of compounds with high p-Ser65-Ub response, thus leaving approximately 1,300 hit compounds (Fig. 4D-F). Our third counter-screening parameter was a p-Ser65-Ub assay without a mitophagy-inducing stimulus (without AA/O). Approximately forty additional compounds which passed the first two filters (Fig. 4F) were eliminated. These compounds were not deemed useful according to our current screening strategy, however they may serve as tool compounds or be potentially useful for other approaches due to their ability to enhance p-Ser65-Ub in the absence of mitochondrial stress. Finally, we applied a filter based on Hoechst fluorescence, excluding compounds which increased or decreased Hoechst fluorescence intensity by greater than two standard deviations from the Hoechst mean intensity of all compounds (Fig. 4E-F). The Hoechst filter excluded compounds which were acutely toxic, auto-fluorescent, or which, through unknown mechanisms, affected Hoechst distribution. In summary, over four hundred 384-well plates were processed, and 123,450 compounds tested in the p-Ser65-Ub ICC phenotypic screen, with 15,197 identified hits, accounting for a primary hit rate of 12.3%. Stringent counter-screening and hit confirmation yielded a shortlist of 1,271 compounds and a final hit rate of 1.03% (Fig. 4F).

      3.5 Chemical and pharmacological profiles of hit compounds and determination of genotype- and stimulus-dependent compound activity

      The 1,271 selected hit compounds were assessed for drug-likeness (Fig. 5A-D). Molecular weight (Fig. 5A), number of hydrogen bond donors (Fig. 5B), LogP (Fig. 5C) and the high percentage passing Lipinski's rule of 5 (Fig. 5D), all suggest that these compounds represent good starting points for further characterization and hit-to-lead optimization in a central nervous system (CNS)-targeting drug discovery program [
      • Lipinski C.A.
      • Lombardo F.
      • Dominy B.W.
      • Feeney P.J.
      Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings.
      ,
      • Wager T.T.
      • Hou X.
      • Verhoest P.R.
      • Villalobos A.
      Central nervous system multiparameter optimization desirability: application in drug discovery.
      ,
      • Rankovic Z.
      CNS drug design: balancing physicochemical properties for optimal brain exposure.
      ].
      Fig 5
      Fig. 5Chemical and pharmacological profiles of hit compounds and determination of genotype- and stimulus-dependent compound activity. A, Distribution of molecular weight of 1,271 hit compounds. B, Distribution of number of hydrogen bond donors of 1,271 hit compounds. C, Distribution of predicted partition coefficient (LogP) of 1,271 hit compounds, D, Distribution of 1,271 hit compounds which pass or fail Lipinski's Rule of 5. E, Representative p-Ser65-Ub concentration-responses, Emax (maximal response within tested concentration range), and pEC50 (negative logarithm of the half maximal effective concentration), over five-point concentration curves (1:3 dilutions, 60 µM to 0.6 µM concentration range) using HCS conditions in Parkin +/R275W fibroblasts. F, Representative images (p-Ser65-Ub in red, Hoechst in blue) corresponding to compound concentration-response curves in E. G, p-Ser65-Ub response (p-Ser65-Ub intensity/nuclei, raw values) in Parkin mutant (Parkin +/R275W) and control (Parkin +/+ (1)) human fibroblasts treated with DMSO (negative control) or FCCP (20 µM; positive control) under optimized HCS assay conditions (AA/O stimulus). H, p-Ser65-Ub response in Parkin mutant (Parkin +/R275W) and control (Parkin +/+ (1)) human fibroblasts treated with DMSO (negative control) or FCCP (20 µM; positive control) using 2 h treatment with 2 µM FCCP as the stimulus (optimized HCS assay conditions for all other factors). I, Scatter plot and straight line fit of p-Ser65-Ub normalized response to test compounds using AA/O stimulus (HCS assay conditions), in Parkin mutant (Parkin +/R275W; x-axis) and control (Parkin +/+ (1); y-axis) human fibroblasts. J, Scatter plot of p-Ser65-Ub normalized response to test compounds with 2 µM FCCP as the stimulus, in Parkin mutant (Parkin +/R275W; x-axis) and control (Parkin +/+ (1); y-axis) human fibroblasts. For G-H, data are mean ± s.d. (n=96 (DMSO conditions) and n=48 (positive control) technical replicates). For I-J, grey stars indicate PPA-1 and number of tested compounds is 960.
      To allow us to rank our hit compounds and further explore their effect on p-Ser65-Ub, we determined concentration-response relationships in the p-Ser65-Ub assay (Fig. 5E-F). Diverse concentration response profiles were present (examples in Fig. 5E-F). Maximal response ranged from 15.3 to 125.0 % normalized p-Ser65-Ub response, with the majority between 20 and 60%. Most compounds (about 75%) had a pEC50 between 4 and 5, representing an EC50 in the medium to high micromolar range, and about 10% of compounds had a pEC50 above 5, in the low micromolar range. Among the examples, ER-000439870 and ER-000482280 had high maximal responses, although ER-000482280 did not reach a plateau over the tested concentration range. ER-001020470, ER-000184896, and ER-001225003 had lower maximal responses, though ER-001020470 and ER-001225003 had high pEC50. We ranked compounds based on pEC50 for further investigation, to identify potential potent compounds of interest. While maximal response is also important, we have observed that high p-Ser65-Ub responses can correlate with MMP depolarization (Fig. 4D), so we prioritized pEC50 to reduce the impact of depolarizing compounds.
      To further confirm compound activity, we tested selected hits in fibroblasts from a healthy control subject expressing common variant Parkin (PRKN +/+; Fig. 5G-J), comparing an alternative mitophagy-inducing stimulus in place of AA/O, selecting the commonly used protonophore and mitochondrial membrane uncoupler FCCP at 2 µM (Fig. 5H and J). In the absence of test compound, the p-Ser65-Ub response to AA/O alone (0.5 µM AA) was approximately half in the Parkin +/R275W compared to Parkin +/+ fibroblasts (negative control; Fig. 5G). This genotype relationship was maintained following incubation with the positive control, 20 µM FCCP (Fig. 5G). Under mild FCCP (2 µM)-mediated mitochondrial stress alone, as well as in combination with the positive control (FCCP at 20 µM), the same genotype effect was also observed (Fig. 5H). Interestingly, p-Ser65-Ub intensity with 2 µM FCCP alone was reduced compared to 0.5 µM AA/O alone, suggesting that at these concentrations FCCP was less capable of inducing p-Ser65-Ub accumulation (Fig. 5G-H). However, addition of the positive control (FCCP at 20 µM) yielded a similar genotype dependent p-Ser65-Ub response, independent of initial stimulus (Fig. 5G-H), suggesting maximal pathway activation.
      In the presence of AA/O-induced mitochondrial stress (HCS assay conditions), compound activity based on normalized p-Ser65-Ub response correlated well between Parkin +/R275W and Parkin +/+ fibroblasts (Fig. 5I). The correlation indicates that the compound activity as a proportion of the positive control is relatively conserved and independent of genotype (Fig. 5I). In contrast, when using FCCP as the mitophagy stimulus instead of AA/O, hit compounds behaved differently between Parkin genotypes (Fig. 5J). Considerably fewer compounds hit overall, and the normalized p-Ser65-Ub response was dissimilar between the two cell lines. Instead, only compounds with >50% p-Ser65-Ub response in the control cells were also active in the Parkin +/R275W cells (Fig. 5J), suggesting a genotype dependency in these conditions. Despite the weak correlation across genotypes under mild FCCP observed for the test compounds, PPA-1 worked equally well between the two cell lines (grey stars in Fig. 5I-J).

      3.6 Functional and mechanistic insights into hit compound activity

      p-Ser65-Ub accumulation reflects initiation of mitophagy and may not correlate with en masse mitochondrial degradation via the lysosome. To determine functional effects of our hit compounds on mitophagy, an eGFP clearance assay was established. Mitochondrially targeted-eGFP (mito-eGFP) was expressed transiently using a custom mRNA and expression was detected using high-content fluorescence imaging. Peak eGFP fluorescence occurred 20-24 h post-transfection, following which a decay in eGFP signal intensity was observed. We demonstrated that mitophagy inducers, including AA/O, FCCP and valinomycin, increased the rate of decay of eGFP signal compared to DMSO (Fig. 6A; blue versus black), suggestive of mitochondrial clearance. Notably, co-incubation of bafilomycin A1 (BAF-A1), a vacuolar ATPase inhibitor which disrupts lysosomal function and autophagic flux, prevented the eGFP signal decay (Fig. 6A; pink), suggesting the decline in eGFP fluorescence after mitophagy induction is mediated via lysosomal/mitophagic degradation.
      Fig 6
      Fig. 6Functional and mechanistic insights into hit compound activity. A, Mito-eGFP intensity over a kinetic analysis of 28 h, imaging at 2 h intervals, in the presence of mitophagy inducer (blue lines; 5 µM AA/O, 10 µM FCCP or 100 nM valinomycin). Bafilomycin A1 (BAF-A1, pink line) was included to inhibit lysosomal turnover of mito-eGFP (10 nM; 1 h pre-incubation). B, Representative compound displaying enhanced mito-eGFP clearance. A five-point concentration range of AA/O (0.04-5 µM) was used to induce mitophagy in cells subject to a single concentration of test compound or vehicle control (DMSO). Mito-eGFP fluorescence was recorded in live cells at 28 h. Data normalized to DMSO (0% clearance) and valinomycin (200 nM; 100% clearance) without AA/O. C, Scatter plot of mito-eGFP clearance and TMRM response from 960 compounds, each run in singlicate. Hit compounds were pre-incubated for 2 h at 20 µM before addition of mitophagy inducing-stimulus and 1:2 dilution to a final assay concentration of 10 µM. A five-point concentration range of AA/O (0.04-5 µM) was used to induce mitophagy in cells subject to a single concentration of test compound or vehicle control (DMSO). Mito-eGFP fluorescence was recorded in live cells at 28 h. Area under the five-point AA/O concentration curve (AUC) was calculated as normalized to assay controls valinomycin (200 µM) and DMSO without AA/O (100% and 0% mito-eGFP clearance respectively) and presented as fold change over the average DMSO AUC (derived from 6 technical DMSO replicates) on each assay plate. AUC was calculated based on the area between the fitted curve and the x-axis; test compound AUC is indicated by pale blue fill in B, DMSO AUC is indicated by dark blue check overlay, as indicated in B. Dots are colored based on PINK1 AUC, derived as for mito-eGFP and TMRM. Gradient from low PINK1 (<1.2-fold change versus DMSO) in blue to high PINK (>1.4-fold change versus DMSO) in pink. D, Scatter plot of USP30-ubiquitin rhodamine 110 (Ub-Rho110) screen from ∼1,200 compounds, to identify USP30 inhibitors. Compounds (including USP30Inh-1 positive control) pre-incubated at 20 µM for 30 min, before addition of 2xUb-Rho110 substrate. Normalized USP30 inhibition after 1 h reported (negative control, DMSO, 0% inhibition; positive control, USP30Inh-1, 100% inhibition). E, Summary characteristics of putative USP30 inhibitors over three assays. Grey shading indicates a positive outcome. TMRM and mito-eGFP data derived from C. F, DUB selectivity assay (DUBprofiler™) was conducted using 50 µM of ER-000340242 and ER-000471258. For A, data are mean ± s.d. (n=4 independent experiments). For B, data are mean ± s.d. (n=6 DMSO technical replicates). For F, data are mean of 2 technical replicates.
      We then used our mito-eGFP clearance assay to assess hit compound-mediated mitochondrial clearance. We measured mito-eGFP clearance in live cells after 28 h compound treatment over a range of AA/O stimulus (exemplar compound, Fig. 6B; x-axis, Fig. 6C). A TMRM assay was carried out in the same samples following mito-eGFP acquisition to determine compound effects on MMP (Fig 6C; y-axis). To further multiplex this assay, following TMRM, the plate was processed for PINK1 ICC (Fig 6C; coloured circles). Leveraging the orthogonal read-outs, a range of compound profiles were identified from this single experiment. Interestingly, compounds with high MMP depolarization also had high PINK1 signal, aligning with a known mechanism of PINK1 stabilization and validating this approach. Notably, many compounds produced an increase in eGFP clearance (greater AUC) without perturbing MMP (Fig. 6C), with some even producing PINK1 stabilization without MMP perturbation (Fig. 6C, black arrows).
      Finally, as validation that our HCS and counter-screening approach yielded mechanistically relevant compounds, we searched within our 1,200 hits for inhibitors of the mitophagy negative regulator and deubiquitinase (DUB), USP30 (Fig. 6D). USP30 opposes mitochondrial substrate ubiquitination, modulating the threshold for mitophagy initiation [
      • Rusilowicz-Jones E.V.
      • et al.
      USP30 sets a trigger threshold for PINK1-PARKIN amplification of mitochondrial ubiquitylation.
      ,
      • Bingol B.
      • et al.
      The mitochondrial deubiquitinase USP30 opposes parkin-mediated mitophagy.
      ]. Inhibition of USP30 was assessed biochemically using the artificial fluorogenic DUB substrate, ubiquitin-rhodamine 110 (Ub-Rho110). Six compounds produced >15% inhibition of recombinant USP30 (Fig. 6D). Based on compound characterisation across both the USP30 activity and mito-eGFP functional assays (Fig. 6E), ER-000340242 and ER-000471258 were selected for profiling using the Ubiquigent DUBprofiler™ service. ER-000471258 demonstrated broad inhibitory effects against >40 known DUB enzymes at 50 µM, with very strong activity against USP8. In contrast, ER-000340242 proved more selective, with weak inhibitory effect (20-30% inhibition) only against USP30 (Fig. 6F).

      4. Discussion

      Defective mitochondrial quality control and accumulation of dysfunctional mitochondria are involved in the etiology of PD. Therefore, maintaining a healthy mitochondrial network through accelerating clearance of damaged mitochondria via mitophagy presents a therapeutic opportunity. Given the complexities of cellular signalling events, high-throughput screening (HTS)-compatible, target-based biochemical assays, performed within an isolated or reconstituted system, may not represent the biology in a cellular context. Instead, in this study we have chosen a phenotypic screening approach, measuring p-Ser65-Ub accumulation. p-Ser65-Ub is a key upstream event in mitophagy initiation, reporting on activity of both PINK1 and Parkin, products of two PD-related genes, and is proven to be a key marker of mitochondrial quality control activation in vivo [
      • Fiesel F.C.
      • Springer W.
      Disease relevance of phosphorylated ubiquitin (p-S65-Ub).
      ]. Together, the strong genetic rationale and clear cellular signalling events mediated by PINK1 and Parkin supported the development of an HCS-compatible assay.
      We believe our phenotypic screening strategy has multiple advantages. (1) The HCS is mechanism-of-action and target-agnostic, likely to identify novel small molecule modulators of in-cell mitophagy; (2) the p-Ser65-Ub endpoint is pertinent to human PD genetics and relevant to the biology underlying disease pathophysiology (i.e., cellular ability to clear dysfunctional mitochondria); (3) due to its phenotypic nature it allows for identification of previously unknown molecular regulators of p-Ser65-Ub, not limited to canonical PINK1-Parkin-mediated mitophagy; and (4) it uses a cell system that is not based on Parkin overexpression, making it a more relevant physiological model. Together, the HCS may provide novel starting material for PD drug development, and a potent, selective small molecule therapeutic derived from this approach will likely have utility in sporadic PD.
      To develop a robust high-content phenotypic screen measuring compound-induced changes in p-Ser65-Ub, we selected Parkin+/R275W fibroblasts as a cellular model. These fibroblasts also have several advantages: (1) they manifest clear mitophagy deficits that can potentially be normalized; (2) despite these mitophagy deficits, the residual Parkin activity in these cells leaves them capable of accumulating p-Ser65-Ub (Fig. 1B) and allows the PINK1-Parkin pathway to be pharmacologically modulated (See [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ] and Figs. 1B, 3); (3) fibroblasts are easily imaged and have clearly identifiable reticular mitochondria; and (4) are readily expandable and amenable to HCS.
      Selection of an appropriate cell model was followed by careful optimization and validation of all aspects of the p-Ser65-Ub HCS assay (Fig. 1). We further developed a high-content TMRM assay for measurement of mitochondrial membrane potential (MMP; Fig. 2), to be used in counter-screening to remove compounds detrimentally affecting MMP. A proof-of-concept screen of forty five selected literature compounds demonstrated that several compounds suggested to act via the PINK1-Parkin pathway were also active in our screening system, validating our approach (Fig. 3) [
      • Katayama H.
      • et al.
      Visualizing and modulating mitophagy for therapeutic studies of neurodegeneration.
      ,
      • Moskal N.
      • et al.
      ROCK inhibitors upregulate the neuroprotective Parkin-mediated mitophagy pathway.
      ,
      • Liu Y.
      • et al.
      Chemical inhibition of FBXO7 reduces inflammation and confers neuroprotection by stabilizing the mitochondrial kinase PINK1.
      ,
      • Shiba-Fukushima K.
      • et al.
      A cell-based high-throughput screening identified two compounds that enhance PINK1-Parkin signaling.
      ].
      Over a period of four months, aided by careful selection of the most appropriate assay automation methods, approximately thirty 384-well plates were run on a weekly basis. We successfully screened a defined subset of the Eisai small molecule compound collection and identified molecules with the capacity to enhance p-Ser65-Ub accumulation in Parkin +/R275W patient-derived fibroblasts (Fig. 4). Following rigorous counter-screening, a final hit rate of 1% was determined, corresponding to our prediction in optimization studies (Fig. 2). Hits were pharmacologically profiled across a five-point concentration range allowing further hit confirmation and prioritization, and activity was confirmed in Parkin common variant fibroblasts (Fig. 5). We developed a mito-eGFP clearance assay to report on compound-induced functional clearance of mitochondrially-localized protein. Functional confirmation of hit compound activity was successfully multiplexed with chronic TMRM assessment and determination of compound-mediated PINK1 stabilization: all three endpoints were determined over a concentration range of mitophagy-inducing stressor (Fig. 6). We have used these endpoints to prioritize and further shortlist our hit compounds.
      As an example of a target-based deconvolution approach, we screened our hit compounds for USP30 inhibition, a proposed mechanism for mitophagy enhancement [
      • Bingol B.
      • et al.
      The mitochondrial deubiquitinase USP30 opposes parkin-mediated mitophagy.
      ]. We have previously observed that USP30Inh-1, 2 and 3 increase FCCP-mediated p-Ser65-Ub accumulation [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ], and we further confirmed this in our proof-of-concept screen, observing a concentration-dependent increase in p-Ser65-Ub with these compounds, in the absence of mitochondrial membrane depolarization (Fig. 3). Using an established USP30 biochemical assay [
      • Tsefou E.
      • et al.
      Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
      ], we identified several putative USP30 inhibitors, one of which (ER-000340242) proved selective across a panel of DUB enzymes (Fig. 6). These data provide confirmation that our HCS can yield mechanistically relevant biology and compounds. Further, comprehensive target-based and unbiased deconvolution approaches are required to understand compound mechanism-of-action of remaining hits.
      One interesting observation was the dependence of hit compound activity as determined by the mitophagy-inducing stimuli across PRKN genotypes. We confirmed strong correlation between hit compound activity in PRKN common variant fibroblasts using the HCS mild AA/O stimulus, suggesting no genotype dependency (Fig. 5). However, using FCCP as an alternative mitophagy-inducing stimulus revealed a strong genotype effect. Hit compounds were more able to induce p-Ser65-Ub in PRKN common variant fibroblasts compared to Parkin +/R275W cells. One explanation for this may be that FCCP, at this low concentration, is less able to trigger mitophagy initiation and stabilize PINK1 in Parkin mutant cells. This is supported by the observation that the response to mild AA/O is greater in both cell lines compared to FCCP (Fig. 5G-H). It may be the threshold for mitophagy initiation has not been reached sufficiently in the PRKN mutant cells, given the feed-forward mechanism of Parkin function, and greater time or FCCP concentration is required. The dependence of hit compound activity on the mitophagy-inducing stress used (Fig. 5) emphasizes the future value of disease-relevant in vitro models more closely reflecting pathophysiological stress.
      In summary, we have successfully miniaturized, optimized and completed a phenotypic HCS using the p-Ser65-Ub endpoint in Parkin +/R275W fibroblasts. Through a well-defined workflow, we have identified compounds worthy of further investigation, which show activity in a novel functional assay, and which include modulators of the known mitophagy regulator USP30. Deconvolution and compound assessment in physiological and disease-based assays has the potential to identify novel biological regulators of PINK1-Parkin biology and provide starting material for drug discovery.

      CRediT authorship contribution statement

      Roberta Tufi: Investigation, Data curation, Formal analysis, Visualization, Writing - original draft. Emily H. Clark: Investigation, Data curation, Formal analysis, Visualization, Writing - original draft. Tamaki Hoshikawa: Project administration, Data curation, Visualization, Writing - review & editing. Christiana Tsagkaraki: Investigation, Data curation, Formal analysis, Writing - review & editing. Jack Stanley: Investigation, Data curation. Kunitoshi Takeda: Project administration. James M. Staddon: Funding acquisition, Writing - review & editing. Thomas Briston: Conceptualization, Funding acquisition, Supervision, Investigation, Data curation, Visualization, Formal analysis, Writing - original draft.

      Declaration of competing interests

      The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Thomas Briston reports a relationship with Eisai Ltd that includes: employment. Roberta Tufi reports a relationship with Eisai Ltd that includes: employment. Emily H. Clark reports a relationship with Eisai Ltd that includes: employment. Tamaki Hoshikawa reports a relationship with Eisai Ltd that includes: employment. Christiana Tsagkaraki reports a relationship with Eisai Ltd that includes: employment. Jack Stanley reports a relationship with Eisai Ltd that includes: employment. Kunitoshi Takeda reports a relationship with Eisai Ltd that includes: employment. James M. Staddon reports a relationship with Eisai Ltd that includes: employment.

      Acknowledgments

      We thank Robin Ketteler and Eliona Tsefou (University College London) for their input into the project. We are grateful to the Michael J Fox Foundation (MJFF Grant ID: MJFF-009659) for generously funding a portion of this study. We would like to thank members of the UCL: Eisai Therapeutic Innovation Group (Andy Takle, Tom Warner, Peter Atkinson, Hélène Plun-Favreau, Adrian Isaacs, Anthony Groom, Jane Kinghorn and Nicola Ridgway) for scientific discussions, guidance, and critique of the project. Finally, we thank Aurelio Vázquez de la Torre for his contribution in the early stages of this project.

      Appendix. Supplementary materials

      References

        • Osellame L.D.
        • et al.
        Mitochondria and quality control defects in a mouse model of Gaucher disease-links to Parkinson's disease.
        Cell Metab. 2013; 17: 941-953https://doi.org/10.1016/j.cmet.2013.04.014
        • Valente E.M.
        • et al.
        Hereditary early-onset Parkinson's disease caused by mutations in PINK1.
        Science (New York, N.Y.). 2004; 304: 1158-1160https://doi.org/10.1126/science.1096284
        • Clark I.E.
        • et al.
        Drosophila pink1 is required for mitochondrial function and interacts genetically with Parkin.
        Nature. 2006; 441: 1162-1166https://doi.org/10.1038/nature04779
        • Park J.
        • et al.
        Mitochondrial dysfunction in Drosophila PINK1 mutants is complemented by Parkin.
        Nature. 2006; 441: 1157-1161https://doi.org/10.1038/nature04788
        • Kitada T.
        • et al.
        Mutations in the parkin gene cause autosomal recessive juvenile parkinsonism.
        Nature. 1998; 392: 605-608https://doi.org/10.1038/33416
        • Fiesel F.C.
        • et al.
        Patho-)physiological relevance of PINK1-dependent ubiquitin phosphorylation.
        EMBO Rep. 2015; 16: 1114-1130https://doi.org/10.15252/embr.201540514
        • Hou X.
        • et al.
        Age- and disease-dependent increase of the mitophagy marker phospho-ubiquitin in normal aging and Lewy body disease.
        Autophagy. 2018; 14: 1404-1418https://doi.org/10.1080/15548627.2018.1461294
        • Silvestri L.
        • et al.
        Mitochondrial import and enzymatic activity of PINK1 mutants associated to recessive parkinsonism.
        Hum Mol Genet. 2005; 14: 3477-3492https://doi.org/10.1093/hmg/ddi377
        • Clark E.H.
        • Vázquez de la Torre A.
        • Hoshikawa T.
        • Briston T.
        Targeting mitophagy in Parkinson's disease.
        J Biol Chem. 2021; 296100209https://doi.org/10.1074/jbc.REV120.014294
      1. Padmanabhan S., Polinski N.K., Menalled L.B., Baptista M.A.S., Fiske B.K. The Michael J. Fox Foundation for Parkinson’s research strategy to advance therapeutic development of PINK1 and Parkin. Biomolecules 2019 Jul 24;9(8):296. doi:10.3390/biom9080296.

        • Lesage S.
        • et al.
        Characterization of recessive Parkinson disease in a large multicenter study.
        Ann Neurol. 2020; 88: 843-850https://doi.org/10.1002/ana.25787
        • Narendra D.
        • Tanaka A.
        • Suen D.F.
        • Youle R.J.
        Parkin is recruited selectively to impaired mitochondria and promotes their autophagy.
        J Cell Biol. 2008; 183: 795-803https://doi.org/10.1083/jcb.200809125
        • Keeney P.M.
        • Xie J.
        • Capaldi R.A.
        • Bennett J.P.
        Parkinson's disease brain mitochondrial complex I has oxidatively damaged subunits and is functionally impaired and misassembled.
        J Neurosci. 2006; 26: 5256-5264https://doi.org/10.1523/jneurosci.0984-06.2006
        • Ryan B.J.
        • Hoek S.
        • Fon E.A.
        • Wade-Martins R.
        Mitochondrial dysfunction and mitophagy in Parkinson's: from familial to sporadic disease.
        Trends Biochem Sci. 2015; 40: 200-210https://doi.org/10.1016/j.tibs.2015.02.003
        • Bose A.
        • Beal M.F.
        Mitochondrial dysfunction in Parkinson's disease.
        J Neurochem. 2016; 139: 216-231https://doi.org/10.1111/jnc.13731
      2. Requejo-Aguilar, R. & Bolanos, J. P. Mitochondrial control of cell bioenergetics in Parkinson's disease. Free radical biology & medicine 100, 123-137, doi:10.1016/j.freeradbiomed.2016.04.012 (2016).

        • Luo Y.
        • Hoffer A.
        • Hoffer B.
        • Qi X.
        Mitochondria: a therapeutic target for Parkinson's disease?.
        Int J Mol Sci. 2015; 16: 20704-20730https://doi.org/10.3390/ijms160920704
        • Hsieh C.H.
        • et al.
        Functional impairment in miro degradation and mitophagy is a shared feature in familial and sporadic Parkinson's disease.
        Cell Stem Cell. 2016; 19: 709-724https://doi.org/10.1016/j.stem.2016.08.002
        • Toomey C.E.
        • et al.
        Mitochondrial dysfunction is a key pathological driver of early stage Parkinson's.
        Acta Neuropathol Commun. 2022; 10: 134https://doi.org/10.1186/s40478-022-01424-6
        • Soutar M.P.M.
        • et al.
        Regulation of mitophagy by the NSL complex underlies genetic risk for Parkinson's disease at 16q11.2 and MAPT H1 loci.
        Brain. 2022; https://doi.org/10.1093/brain/awac325
      3. Billingsley K.J., et al. Mitochondria function associated genes contribute to Parkinson’s Disease risk and later age at onset. NPJ Parkinsons Dis 2019 May 22;5:8. doi:10.1038/s41531-019-0080-x.

        • Yamano K.
        • Youle R.J.
        PINK1 is degraded through the N-end rule pathway.
        Autophagy. 2013; 9: 1758-1769https://doi.org/10.4161/auto.24633
        • Narendra D.P.
        • et al.
        PINK1 is selectively stabilized on impaired mitochondria to activate Parkin.
        PLoS Biol. 2010; 8e1000298https://doi.org/10.1371/journal.pbio.1000298
        • Okatsu K.
        • et al.
        A dimeric PINK1-containing complex on depolarized mitochondria stimulates Parkin recruitment.
        J Biol Chem. 2013; 288: 36372-36384https://doi.org/10.1074/jbc.M113.509653
        • Okatsu K.
        • et al.
        PINK1 autophosphorylation upon membrane potential dissipation is essential for Parkin recruitment to damaged mitochondria.
        Nat Commun. 2012; 3: 1016https://doi.org/10.1038/ncomms2016
        • Kondapalli C.
        • et al.
        PINK1 is activated by mitochondrial membrane potential depolarization and stimulates Parkin E3 ligase activity by phosphorylating Serine 65.
        Open Biol. 2012; 2120080https://doi.org/10.1098/rsob.120080
        • Sulkshane P.
        • et al.
        Inhibition of proteasome reveals basal mitochondrial ubiquitination.
        J Proteom. 2020; 229103949https://doi.org/10.1016/j.jprot.2020.103949
        • Kane L.A.
        • et al.
        PINK1 phosphorylates ubiquitin to activate Parkin E3 ubiquitin ligase activity.
        J Cell Biol. 2014; 205: 143-153https://doi.org/10.1083/jcb.201402104
        • Kazlauskaite A.
        • et al.
        Binding to serine 65-phosphorylated ubiquitin primes Parkin for optimal PINK1-dependent phosphorylation and activation.
        EMBO Rep. 2015; 16: 939-954https://doi.org/10.15252/embr.201540352
        • Gladkova C.
        • Maslen S.L.
        • Skehel J.M.
        • Komander D.
        Mechanism of Parkin activation by PINK1.
        Nature. 2018; 559: 410-414https://doi.org/10.1038/s41586-018-0224-x
        • Sarraf S.A.
        • et al.
        Landscape of the PARKIN-dependent ubiquitylome in response to mitochondrial depolarization.
        Nature. 2013; 496: 372-376https://doi.org/10.1038/nature12043
        • Wauer T.
        • Simicek M.
        • Schubert A.
        • Komander D.
        Mechanism of phospho-ubiquitin-induced PARKIN activation.
        Nature. 2015; 524: 370-374https://doi.org/10.1038/nature14879
        • Ordureau A.
        • et al.
        Quantitative proteomics reveal a feedforward mechanism for mitochondrial PARKIN translocation and ubiquitin chain synthesis.
        Mol Cell. 2014; 56: 360-375https://doi.org/10.1016/j.molcel.2014.09.007
        • Tanaka A.
        • et al.
        Proteasome and p97 mediate mitophagy and degradation of mitofusins induced by Parkin.
        J Cell Biol. 2010; 191: 1367-1380https://doi.org/10.1083/jcb.201007013
        • Wang X.
        • et al.
        PINK1 and Parkin target Miro for phosphorylation and degradation to arrest mitochondrial motility.
        Cell. 2011; 147: 893-906https://doi.org/10.1016/j.cell.2011.10.018
        • Lazarou M.
        • et al.
        The ubiquitin kinase PINK1 recruits autophagy receptors to induce mitophagy.
        Nature. 2015; 524: 309-314https://doi.org/10.1038/nature14893
        • Marcassa E.
        • et al.
        Dual role of USP30 in controlling basal pexophagy and mitophagy.
        EMBO Rep. 2018; 19e45595https://doi.org/10.15252/embr.201745595
        • Rusilowicz-Jones E.V.
        • et al.
        USP30 sets a trigger threshold for PINK1-PARKIN amplification of mitochondrial ubiquitylation.
        Life Sci Alliance. 2020; 3https://doi.org/10.26508/lsa.202000768
        • Tsefou E.
        • et al.
        Investigation of USP30 inhibition to enhance Parkin-mediated mitophagy: tools and approaches.
        Biochem J. 2021; 478: 4099-4118https://doi.org/10.1042/bcj20210508
      4. Rusilowicz-Jones E.V., et al. Benchmarking a highly selective USP30 inhibitor for enhancement of mitophagy and pexophagy. Life Sci Alliance 2021 Nov 29;5(2):e202101287. doi:10.26508/lsa.202101287.

        • Bingol B.
        • et al.
        The mitochondrial deubiquitinase USP30 opposes parkin-mediated mitophagy.
        Nature. 2014; 510: 370-375https://doi.org/10.1038/nature13418
        • Wauer T.
        • et al.
        Ubiquitin Ser65 phosphorylation affects ubiquitin structure, chain assembly and hydrolysis.
        EMBO J. 2015; 34: 307-325https://doi.org/10.15252/embj.201489847
        • Ordureau A.
        • et al.
        Global landscape and dynamics of Parkin and USP30-dependent ubiquitylomes in iNeurons during mitophagic signaling.
        Mol Cell. 2020; 77 (.e1110): 1124-1142https://doi.org/10.1016/j.molcel.2019.11.013
        • Schubert A.F.
        • et al.
        Structure of PINK1 in complex with its substrate ubiquitin.
        Nature. 2017; 552: 51https://doi.org/10.1038/nature24645
        • Zhang J.H.
        • Chung T.D.
        • Oldenburg K.R.
        A simple statistical parameter for use in evaluation and validation of high throughput screening assays.
        J Biomol Screen. 1999; 4: 67-73https://doi.org/10.1177/108705719900400206
        • Jin S.M.
        • et al.
        Mitochondrial membrane potential regulates PINK1 import and proteolytic destabilization by PARL.
        J Cell Biol. 2010; 191: 933-942https://doi.org/10.1083/jcb.201008084
        • Marroquin L.D.
        • Hynes J.
        • Dykens J.A.
        • Jamieson J.D.
        • Will Y.
        Circumventing the Crabtree effect: replacing media glucose with galactose increases susceptibility of HepG2 cells to mitochondrial toxicants.
        Toxicol Sci. 2007; 97: 539-547https://doi.org/10.1093/toxsci/kfm052
        • Moffat J.G.
        • Vincent F.
        • Lee J.A.
        • Eder J.
        • Prunotto M.
        Opportunities and challenges in phenotypic drug discovery: an industry perspective.
        Nat Rev Drug Discov. 2017; 16: 531-543https://doi.org/10.1038/nrd.2017.111
        • Katayama H.
        • et al.
        Visualizing and modulating mitophagy for therapeutic studies of neurodegeneration.
        Cell. 2020; 181 (e1116): 1176-1187https://doi.org/10.1016/j.cell.2020.04.025
        • Moskal N.
        • et al.
        ROCK inhibitors upregulate the neuroprotective Parkin-mediated mitophagy pathway.
        Nat Commun. 2020; 11: 88https://doi.org/10.1038/s41467-019-13781-3
        • Liu Y.
        • et al.
        Chemical inhibition of FBXO7 reduces inflammation and confers neuroprotection by stabilizing the mitochondrial kinase PINK1.
        JCI Insight. 2020; 5https://doi.org/10.1172/jci.insight.131834
        • Shiba-Fukushima K.
        • et al.
        A cell-based high-throughput screening identified two compounds that enhance PINK1-Parkin signaling.
        iScience. 2020; 23101048https://doi.org/10.1016/j.isci.2020.101048
        • Osgerby L.
        • et al.
        Kinetin riboside and its ProTides activate the Parkinson's disease associated PTEN-induced putative kinase 1 (PINK1) independent of mitochondrial depolarization.
        J Med Chem. 2017; 60: 3518-3524https://doi.org/10.1021/acs.jmedchem.6b01897
        • Cerqueira F.M.
        • et al.
        MitoTimer-based high-content screen identifies two chemically-related benzothiophene derivatives that enhance basal mitophagy.
        Biochem J. 2020; 477: 461-475https://doi.org/10.1042/bcj20190616
        • East D.A.
        • et al.
        PMI: a ΔΨm independent pharmacological regulator of mitophagy.
        Chem Biol. 2014; 21: 1585-1596https://doi.org/10.1016/j.chembiol.2014.09.019
        • Allen G.F.
        • Toth R.
        • James J.
        • Ganley I.G.
        Loss of iron triggers PINK1/Parkin-independent mitophagy.
        EMBO Rep. 2013; 14: 1127-1135https://doi.org/10.1038/embor.2013.168
        • Karuppagounder S.S.
        • et al.
        The c-Abl inhibitor, nilotinib, protects dopaminergic neurons in a preclinical animal model of Parkinson's disease.
        Sci Rep. 2014; 4: 4874https://doi.org/10.1038/srep04874
        • Hsieh C.H.
        • et al.
        Miro1 marks Parkinson's disease subset and Miro1 reducer rescues neuron loss in Parkinson's models.
        Cell Metab. 2019; 30 (e1137): 1131-1140https://doi.org/10.1016/j.cmet.2019.08.023
        • Lipinski C.A.
        • Lombardo F.
        • Dominy B.W.
        • Feeney P.J.
        Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings.
        Adv Drug Deliv Rev. 2001; 46: 3-26https://doi.org/10.1016/s0169-409x(00)00129-0
        • Wager T.T.
        • Hou X.
        • Verhoest P.R.
        • Villalobos A.
        Central nervous system multiparameter optimization desirability: application in drug discovery.
        ACS Chem Neurosci. 2016; 7: 767-775https://doi.org/10.1021/acschemneuro.6b00029
        • Rankovic Z.
        CNS drug design: balancing physicochemical properties for optimal brain exposure.
        J Med Chem. 2015; 58: 2584-2608https://doi.org/10.1021/jm501535r
        • Fiesel F.C.
        • Springer W.
        Disease relevance of phosphorylated ubiquitin (p-S65-Ub).
        Autophagy. 2015; 11: 2125-2126https://doi.org/10.1080/15548627.2015.1091912