Ligand-free mitochondria-localized mutant AR-induced cytotoxicity in spinal bulbar muscular atrophy

Abstract

Spinal bulbar muscular atrophy (SBMA), the first identified CAG-repeat expansion disorder, is an X-linked neuromuscular disorder involving CAG-repeat-expansion mutations in the androgen receptor (AR) gene. We utilized CRISPR-Cas9 gene editing to engineer novel isogenic human induced pluripotent stem cell (hiPSC) models, consisting of isogenic AR knockout, control and disease lines expressing mutant AR with distinct repeat lengths, as well as control and disease lines expressing FLAG-tagged wild-type and mutant AR, respectively. Adapting a small-molecule cocktail-directed approach, we differentiate the isogenic hiPSC models into motor neuron-like cells with a highly enriched population to uncover cell-type-specific mechanisms underlying SBMA and to distinguish gain- from loss-of-function properties of mutant AR in disease motor neurons. We demonstrate that ligand-free mutant AR causes drastic mitochondrial dysfunction in neurites of differentiated disease motor neurons due to gain-of-function mechanisms and such cytotoxicity can be amplified upon ligand (androgens) treatment. We further show that aberrant interaction between ligand-free, mitochondria-localized mutant AR and F-ATP synthase is associated with compromised mitochondrial respiration and multiple other mitochondrial impairments. These findings counter the established notion that androgens are requisite for mutant AR-induced cytotoxicity in SBMA, reveal a compelling mechanistic link between ligand-free mutant AR, F-ATP synthase and mitochondrial dysfunction, and provide innovative insights into motor neuron-specific therapeutic interventions for SBMA.

ligand-free mutant AR, mitochondrial dysfunction, F-ATP synthase, motor neuron degeneration

Introduction

Spinal bulbar muscular atrophy (SBMA) is caused by a CAG repeat expansion (n ≥ 38) in the androgen receptor (AR) gene, which encodes an expanded polyglutamine (polyQ) tract in the protein.¹ Both loss- and gain-offunction of mutant AR contributes to motor neuron degeneration and muscle atrophy.² SBMA exhibits full penetrance only in males, consistent with observations from SBMA mouse models that androgens are critical for disease development in conjunction with mutant AR expression.³ Although castration and androgen deprivation are sufficient to reverse disease phenotypes in mouse models, clinical trials using anti-androgen therapy have failed to show notable effects on swallowing function and progression of muscle weakness in human patients.^4,5 Therefore, in addition to the presence of an androgen ligand in concert with mutant AR, other factors might contribute to disease pathogenesis.

Mitochondrial dysfunction occurs in many neurodegenerative disorders. Considering the intense bioenergetic demand and limited regenerative capacity of neurons, mitochondrial dysfunction can have detrimental effects on neuronal survival. Some pathologic proteins, such as amyloid-β, alpha-synuclein, parkin, PINK1, huntingtin and SOD1, undermine mitochondrial function either via association with the organelle and/or interactions with mitochondrial proteins.⁶ Thus far, compelling evidence indicates that mitochondrial dysfunction could serve as a causative factor that contributes to disease pathogenesis rather than merely being a consequence of neurodegeneration.⁷ In the specific case of SBMA, studies have reported mitochondrial abnormalities in disease cells, mouse models and human patients.⁸ Mutant AR interacts with an inner mitochondrial membrane protein in vitro and possibly localizes to mitochondria,^9–12 indicating that the mutant protein may impair mitochondrial function in SBMA through association with mitochondria or even after entering the organelle, executing its ‘toxic’ function via aberrant interactions with other mitochondrial proteins.

To investigate cell-type-specific molecular mechanisms underlying SBMA, we utilized the CRISPR-Cas9 system to generate a series of new isogenic human-induced pluripotent stem cell (hiPSC) models. Upon motor neuron differentiation, multiple mitochondrial impairments occurred within neurites of motor neuron-like cells by introducing a pathological polyQ length into wild-type AR, while these aberrations were rescued by correcting the expanded repeats in mutant AR. Although androgens are widely believed to play an essential role in mutant AR-induced cytotoxicity in SBMA, we found here that, surprisingly, mutant AR protein is sufficient to lead to mitochondrial dysfunction in disease motor neurons in the absence of ligand binding, whereas ligand treatment merely aggravated such impairments. Furthermore, a preferential interaction between ligand-free mutant AR and mitochondrial F-ATP synthase revealed a pathological role for mutant AR protein towards F-ATP synthase function, resulting in significant impairment in mitochondrial respiration, which is a critical function of mitochondria as powerhouses of the cell. Our findings point to vital contributions of mitochondrial dysfunction, mechanistically involving a preferential interaction between ligand-free mutant AR and F-ATP synthase, as a causative factor for motor neuron degeneration in SBMA.

Materials and methods

Plasmids

For AR gene editing, the pCAG-eCas9-GFP-U6-gRNA plasmid (Addgene #79145) was created by subcloning the eSpCas9 sequence from the plasmid eSpCas9 (1.1) (Addgene #71814¹³) into the vector pCAG-SpCas9-GFP-U6-gRNA (Addgene #79144¹⁴) using EcoRV/FseI cloning sites. The plasmid pJZ155-U6-gRNA-BFP was created by subcloning the synthesized DNA fragment mPGK-TagBFP-bGHpolyA into the vector pAC155-pCR8-sgExpression (Addgene #49045¹⁵) using the PstI site. The donor plasmids pUC19-AR23.HDR and pUC19-AR54.HDR were created by subcloning the synthesized DNA fragments, harbouring 23 and 54 repeats in the codon-optimized region flanked by long homologous arms that are complementary to the modification sites in the human AR gene, into the vector pUC19 using the EcoRI/BamHI sites. The donor plasmids pUC19-AR40.HDR and pUC19-AR68.HDR were created by subcloning the synthesized DNA fragments harbouring 40 and 68 repeats into the plasmid pUC19-AR23.HDR using the EagI/ApaI sites. For N-terminal tagging, the donor plasmids pUC19-FLAG.SPOT.AR23.HDR and pUC19-FLAG.SPOT.AR68.HDR were created by subcloning the synthesized DNA fragments harbouring 3×FLAG and SPOT tags as well as linker sequences into the two related donor vectors using the DraIII/EagI sites. All synthesized DNA fragments were purchased from GenScript, Thermo Fisher Scientific and Integrated DNA Technologies. For stable line construction, the plasmids TRE3G-AR23.FLAG.SPOT and TRE3G-AR68.FLAG.SPOT were created by subcloning full-length AR with 23 or 68 repeats into the plasmid TRE3G-EGFP-mAID-exon1 Q18 (a gift from Nelson Cole, NINDS) using the AflII/BstBI sites.

Antibodies and other reagents

Sources for the antibodies used are as follows: AR (H280, rabbit, sc-13062, Santa Cruz), AR (N-20, rabbit, sc-816, Santa Cruz), AR (mouse, 554225, BD), FLAG (mouse, F1804, Sigma-Aldrich), alpha tubulin (mouse, T6199, Sigma-Aldrich), caspase 3 (rabbit, 9662, Cell Signaling), cleaved caspase 3 (rabbit, 9664, Cell Signaling), ATP5B (rabbit, 17247-1-AP, Proteintech), ATP5H (rabbit, 17589-1-AP, Proteintech), ATP5A1 (rabbit, 14676-1-AP, Proteintech), TOM20 (rabbit, sc-11415, Santa Cruz), Oct4 (mouse, 653702, BioLegend), Nanog (rabbit, ab21624, Abcam), ChAT (goat, AB144P, Sigma-Aldrich), TUBB3 (rabbit, 845502, BioLegend), TUBB3 (mouse, 801202, BioLegend), Hb9 (mouse, 81.5C10, DSHB); Alexa Fluor 488-, 555-, 633-conjugated secondary antibodies (Thermo Fisher Scientific), mouse IgG- or rabbit IgG-HRP (Thermo Fisher Scientific) and normal mouse or rabbit IgG (Thermo Fisher Scientific).

Cell culture-related media and other reagents were from Thermo Fisher Scientific, including Essential 8, Essential 8 Flex, DMEM/F12, neurobasal, McCoy’s, Accutase, Dispase II, Glutamax, penicillin/streptomycin, N2 and B27. Sources for molecular biology-related reagents were as follows: restriction endonucleases (New England Biolabs); CRISPR-related enzymes including T7 endonuclease I, T4 PNK, T7 DNA ligase (New England Biolabs), FastDigest BbsI (Thermo Fisher Scientific) and Plasmid-Safe DNase (Lucigen); PCR polymerase including iProof (Bio-Rad) and PfuUltra (Agilent). All other chemicals were from Millipore Sigma unless noted otherwise.

Development of isogenic hiPSC models using CRISPR and the off-target evaluation

The two parental hiPSC lines HT-AR23 and SB-AR68, which are the iPSC lines HT-180C and SB18 expressing AR23CAG and AR68CAG, respectively,^16,17 were used to develop the isogenic hiPSC models of SBMA. To perform AR gene editing while minimizing off-target effects, we used a variant of Cas9 nuclease, eSpCas9 1.1, due to its high on-target specificity and robust cleavage efficiency.¹³ An online CRISPR design tool (Benchling) was used to design guide RNAs (gRNAs) when considering off-target effects.¹⁸ Two sets of gRNAs (gRNA1 or gRNA8, and gRNA12, oligo information is listed in Supplementary Table 1) were used to simultaneously target 5′- and 3′-CAG repeat regions, respectively, in the AR exon1 in order to remove the entire CAG repeat. The repair donor plasmids used for homology-directed DNA repair harbour desired repeat lengths flanked by left and right homologous arms (∼760 and 830 n.t. for LHA and RHA, respectively) that are complementary to the modification sites in the human AR gene. Given possible contributions of pure CAG repeat-containing transcripts to the gained cytotoxicity,¹⁹ we chose to replace the CAG-repeat region in the endogenous AR allele with desired numbers of (CAA/G)_n repeats to assess cytotoxicity only due to a pathological polyQ tract in the mutant protein. Of note, the average CAG repeat number in wild-type AR is 21 ± 3 in healthy individuals, while the repeat size in mutant AR was within the range of 38–68 reported in SBMA patients to date.^17,20

To make AR knockout lines, the plasmid pCAG-eCas9-GFP-U6-gRNA harbouring either gRNA1 or gRNA8 was transfected into the parental hiPSC lines using a lipid-based reagent Lipofectamine Stem (Thermo Fisher Scientific). To edit the CAG-repeats in the endogenous AR gene, three plasmids including pCAG-eCas9-GFP-U6-gRNA1, pJZ155-U6-gRNA12-BFP and a donor with desired repeat length were co-transfected into the parental hiPSC cells. At 24 h post-transfection, cells expressing EGFP (for AR knockout) or both EGFP and BFP (for CAG-repeat replacement) were sorted using fluorescence-activated cell sorting (BD Biosciences) and cultured as single-cell clones for picking. Targeted clonal lines with defined modification were screened using chromosome-targeted and repeat-region PCR followed by DNA sequencing (primer information was listed in Supplementary Table). The final clonal lines were subjected to iPSC identity and purity examination as well as the G-banded karyotyping analysis (WiCell). At least three clones for each isogenic line were obtained.

Off-target effects due to non-specific nuclease-mediated cleavage during CRISPR genome editing were evaluated using a target-PCR-based approach followed by DNA sequencing. The 50 predicted off-target sites were analysed for each gRNA, 15 of which in exon or exon–intron junction regions were amplified using target-specific PCR followed by DNA sequencing for each isogenic line/clone.

Motor neuron differentiation

To differentiate hiPSCs into spinal motor neuron-like cells, a small molecule cocktail-directed approach was adapted.²¹ In brief, hiPSCs were maintained as feeder-free culture. One day before differentiation, hiPSCs were passaged using Dispase (1 U/ml) and cultured in Essential 8 medium supplemented with Y-27632 ROCK inhibitor (10 µM, APExBIO). On Day 0, the medium was changed to neuroepithelia progenitor (NEP) medium, containing the neural medium [DMEM/F12 and neurobasal medium at 1:1, N2 (0.5×), B27 (0.5×), ascorbic acid (0.1 mM, StemCell Technologies, Glutamax (1×) and penicillin/streptomycin (1×)] supplemented with CHIR99021 (CHIR, 3 µM, APExBIO), DMH-1 (DMH, 2 µM, APExBIO) and SB431542 (SB, 2 µM, APExBIO). On Day 6, NEPs were dissociated with Dispase and cultured in motor neuron progenitor (MNP) medium, containing the neural medium supplemented with CHIR (1 µM), DMH (2 µM), SB (2 µM), retinoid acid (RA, 0.1 µM, Stemgent), and purmorphamine (Pur, 0.5 µM, APExBIO). On Day 12, MNPs were dissociated with Dispase and cultured as motor neuron spheres in the neural medium supplemented with RA (0.5 µM) and Pur (1 µM). On Day 20, the spheres were dissociated with Accutase into single cells, which were plated onto poly-D-lysine- and laminin-coated dishes as monolayer culture. Cells were cultured in regular motor neuron medium, containing the neural medium supplemented with RA (0.5 µM), Pur (1 µM), and Compound E (CompE, 0.1 µM, Millipore Sigma) for 4 days. On Day 24, the medium was replaced by progesterone- and corticosterone-free (P4/CORT-free) motor neuron medium. The P4-free supplement N2 and the P4/CORT-free supplement B27 were prepared without adding progesterone and corticosterone. After a 2-day clearance, on Day 26, cells were treated with either vehicle EtOH (1:1000) or ligand R1881 (10 nM, Millipore Sigma) for 8 days. During the differentiation, the medium was changed every 2 days. Motor neuron identity was determined by immunofluorescence staining of a neuronal marker β-tubulin (TUBB3) as well as two spinal motor neuron markers choline acetyltransferase (ChAT) and homeobox Hb9. Differentiation efficiency was calculated based on the percentage of Hb9⁺/TUBB3⁺/ChAT⁺ cells in the whole cell population. Total cell numbers were determined by DAPI staining. Twenty images for each isogenic line were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Measurement of mitochondrial membrane potential

Differentiated motor neuron-like cells were stained sequentially by tetramethylrhodamine ethylester (TMRE; 25 nM, Thermo Fisher Scientific) and MitoTracker Green (10 nM, Thermo Fisher Scientific) at 37°C for 30 min, washed three times with culture medium, then changed to phenol red-free medium before imaging. Confocal images were acquired using a laser scanning confocal microscope equipped with a ×63/1.4 NA oil immersion objective lens (LSM880, Zeiss). Mitochondrial membrane potential (Δψ_m) was assessed by the ratio of TMRE to MitoTracker fluorescence integrated density of mitochondria in neurites of each isogenic line using ImageJ, and then normalized by the isogenic control in the absence of ligand. At least 20 images for each isogenic line in every experiment were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Measurement of mitochondrial oxidative stress

Differentiated motor neuron-like cells were stained sequentially by CellRox Green (2.5 µM, Thermo Fisher Scientific) and MitoTracker Deep Red (50 nM, Thermo Fisher Scientific) at 37°C for 30 min, washed three times with culture medium, then changed to phenol red-free medium before imaging. Confocal images were acquired using confocal LSM880 equipped with a ×63 objective lens. The reactive oxygen species (ROS) level was assessed by the ratio of CellRox to MitoTracker fluorescence integrated density of mitochondria in neurites of each isogenic line using ImageJ, and then normalized by the isogenic control in the absence of ligand. At least 20 images for each isogenic line in every experiment were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Measurement of mitochondrial size

Differentiated motor neuron-like cells were stained by MitoTracker green (10 nM) at 37°C for 30 min, washed three times with culture medium, then changed to phenol red-free medium before imaging. Confocal images were acquired using LSM880 equipped with a ×63 objective lens. The mitochondrial size was assessed by the aspect ratio of the MitoTracker fluorescence integrated density of each individual mitochondrion in neurites of each isogenic line using ImageJ particle analysis. At least 20 images for each isogenic line in every experiment were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Transmission electron microscopy and cristae architecture analysis

Ultrastructure of mitochondria in neurites of differentiated motor neuron-like cells was examined by negative staining and transmission electron microscopy. Briefly, cells were fixed (4% glutaraldehyde in 0.1 N cacodylate buffer, pH 7.4) at room temperature for 30 min followed by an overnight incubation at 4°C. Next, samples were treated with 1% osmium tetroxide, en bloc stained with 1% uranyl acetate, dehydrated through a series of graded ethanol washes and embedded in epoxy resins. Thin sections were poststained with uranyl acetate and lead citrate, examined using a JEOL 200 CX Transmission Electron Microscope and analysed using ImageJ. Cristae numbers per unit length (µm) of mitochondria were assessed for individual mitochondrion in the neurites. Long versus short cristae were defined based on the size of a crista, which is longer or shorter than half of the mitochondrial width, respectively. At least 50 images for each isogenic line were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Seahorse XF Cell Mito Stress Test

Differentiated cells (150 000) were seeded on 96-well plates at the motor neuron monolayer stage, treated with either vehicle or R1881 for 8 days and then subjected to the Seahorse XF Cell Mito Stress Test. Distinct mitochondrial modulators, including oligomycin (1 µM), carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP; 1 µM), rotenone (0.5 µM) and antimycin A (0.5 µM) were used in the assay. Oxygen consumption rate (OCR) was measured using the Seahorse XF96 Analyzer (Agilent) and normalized by protein concentration, which was measured using the bicinchoninic acid (BCA) assay (Thermo Fisher Scientific) with bovine serum albumin (BSA) as a standard. Three key parameters, including basal respiration, ATP-linked respiration and spare respiratory capacity, were determined using corrected OCR values.

Preparation of cell lysates, co-immunoprecipitation, protein gel electrophoresis and immunoblotting analysis

To prepare whole-cell lysates, cells were lysed in cold Triton lysis buffer [pH 7.5, 50 mM Tris–HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, proteinase/phosphatase inhibitor cocktail (1×, Thermo Fisher Scientific) and 1 mM PMSF]. To prepare cell lysates for co-immunoprecipitation (co-IP), cells were lysed in cold NP-40 lysis buffer [pH 7.4, 25 mM Tris–HCl, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 5% glycerol, proteinase/phosphatase inhibitor cocktail (1×) and 1 mM PMSF]. Protein concentrations in cell lysates were determined using the BCA assay.

For co-IP, antibody-conjugated magnetic beads were used. Specifically, anti-AR antibodies were conjugated onto the protein A/G magnetic beads (Thermo Fisher Scientific) for AR co-IP and FLAG M2 magnetic beads (Millipore Sigma) were used for FLAG co-IP. Cells without AR expression, e.g. isogenic AR knockout cells or U2OS cells, were used as a negative control. In brief, aliquots of cell lysates were precleared with IgG controls to remove Ab non-specific binding. Protein–protein interaction complexes were captured by incubating with antibody-conjugated beads at 4°C and eluted using acidic glycine solution (pH 2.0, for AR co-IP) or FLAG peptide (APExBIO, for FLAG co-IP).

Samples for sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) were prepared in lithium dodecyl sulphate sample buffer (Thermo Fisher Scientific) in the presence of beta-mercaptoethanol. Aliquots of cell lysates or co-IP eluates were resolved by SDS-PAGE, transferred onto supported nitrocellulose membranes (0.2 µm, Bio-Rad) and immunoblotted with indicated primary antibodies followed by HRP-conjugated secondary antibodies. Immunoblotting signals were visualized using Pierce ECL (Thermo Fisher Scientific) or Amersham ECL Select western blotting detection reagents (Cytiva), detected using the Odyssey Fc Imaging System (LI-COR), CL-XPosure films (Thermo Fisher Scientific) and Amersham Hyperfilms (Cytiva), and analysed using ImageJ and a densitometry-based method.

Proximity ligation assay

To assess the mitochondrial localization of AR and its interaction with F-ATP synthase, the two binding parties were labelled using anti-FLAG (to detect FLAG-tagged AR) and anti-ATP5A1 antibodies, respectively. The mitochondrial outer membrane protein TOM20 was used to mark the mitochondria fluorescently labelled using the Zenon IgG labelling kit (Thermo Fisher Scientific). In brief, cells were fixed and permeabilized using the indirect immunofluorescence staining protocol (listed below) and then subjected to the proximity ligation assay (PLA; Millipore Sigma) according to the manufacturer’s instructions. Images were taken using a Zeiss LSM880 with Airyscan module equipped with a ×63/1.4 NA oil immersion objective lens. Numbers as well as aspect ratios of PLA-positive puncta per frame were quantitated using the ImageJ particle analysis. At least 20 images for each isogenic line in every experiment were randomly selected and subjected to quantitation, which was obtained from a blinded assessment.

Immunofluorescence and phase-contrast microscopy

For indirect immunofluorescence, cells were fixed in 4% paraformaldehyde, permeabilized using Dulbecco’s phosphate-buffered saline (DPBS) with 0.2% Triton, blocked in 10% BSA in DPBS with 0.1% Triton X-100 and 0.1% Tween 20, incubated sequentially in primary and secondary antibody solutions, counterstained with DAPI and mounted in mounting medium. Fluorescence images were acquired as single focal planes using a Zeiss LSM880 confocal microscope equipped with a ×20/0.80 NA PlanApo objective lens. Phase-contrast images were acquired using an EVOS M5000 imaging system equipped with a ×10 long working distance objective lens (Thermo Fisher Scientific).

Label-free mass spectrometry and gene ontology enrichment analyses

AR co-IP samples were eluted in glycine buffer (pH 2.0). Samples were reduced with 10 mM Tris (2-carboxyethyl) phosphine for 1 h, alkylated with 20 mM N-ethylmaleimide (NEM) for 15 min and digested with trypsin at 37°C overnight. Peptides were desalted using Oasis HLB µElution plates (Waters). Liquid chromatography–tandem mass spectrometry (LC-MS/MS) experiments were performed on a system where an Ultimate 3000 RSLC nano high-performance liquid chromatography unit (Thermo Fisher Scientific) was coupled to an Orbitrap Lumos mass spectrometer (Thermo Fisher Scientific). Peptides were separated on an ES802 column over a 63-min gradient with mobile phase B (98% acetonitrile, 1.9% H₂O, 0.1% formic acid) increased from 3 to 21%. LC-MS/MS data were acquired in data-dependent acquisition mode. MS1 scans were performed in Orbitrap with a resolution of 120 K, a mass range of 375–1500 m/z and an AGC target of 2 × 10⁵. The quadrupole isolation window is 1.5 m/z. The precursor intensity threshold for MS/MS scan was set at 1 × 10⁴. MS2 scans were conducted in ion trap. Peptides were fragmented with the higher-energy C-trap dissociation (HCD) method and the collision energy was fixed at 30%. MS1 scan was performed every 3 s. As many MS2 scans were acquired within the 3-s cycle.

Proteome Discoverer 2.4 was used for database search and protein abundance calculation. LC-MS/MS data were searched against Sprot Human database using the Mascot search engine. Up to one missed cleavage was allowed for trypsin digestion. NEM on cysteines was set as static modification. Oxidation on methionine and acetyl on protein N-termini were set as dynamic modifications. Mass tolerances for MS1 and MS2 scans were set to 5 ppm and 0.25 Da, respectively. Percolator was used for peptide spectrum match validation. The search results were filtered by a false discovery rate of 1% at the protein level. Protein abundance values were calculated for all proteins identified by summing the abundance of unique peptides matched to that protein. Abundances among different conditions were normalized. Protein ratios were calculated by directly comparing the normalized protein abundances between two conditions. P-values were not calculated because there was one sample for each condition.

To compare affinity of candidate binding partners for AR, fold change was used. Fold change > 1.4 was considered as a preferential binding partner to the mutant AR compared with the wild-type AR, whereas fold change < 0.714 was considered as a weaker binding partner to the mutant protein. A list of candidate stronger binding partners to the mutant AR were further subjected to the Gene Ontology (GO)-term identification (GO biological process terms) and enrichment analysis using the online Panther tool.

Development of a FLAG.AR isogenic hiPSC model

A similar strategy of CRISPR genome editing was used to develop FLAG.AR isogenic hiPSC lines (under the Institutional Review Board protocol #00275993 approved by Johns Hopkins University, Office of Human Subjects Research Institutional Review Boards). In brief, three plasmids including pCAG-eCas9-GFP-U6-gRNA1, pJZ155-U6-gRNA12-BFP and a donor expressing 3×FLAG.SPOT-tagged AR with 23 or 68 repeats were co-transfected into the parental hiPSC cells. Targeted clonal lines were sorted, cultured and identified using chromosome-targeted, repeat-region (including the 3×FLAG.SPOT-tag sequence) PCR followed by DNA sequencing. The final clonal lines were subjected to iPSC identity and purity examination, G-banded karyotyping analysis and off-target evaluation. At least two clones for each isogenic line were obtained.

Development of an inducible cell model

The U2OS cell line was chosen to develop an inducible cell model owing to no endogenous expression of wild-type AR and highly amenable to imaging given the flat morphology with nicely distributed mitochondria and other organelles. U2OS cells were transfected with a Tet-inducible plasmid TRE3G-AR23.FLAG.SPOT (expressing full-length AR23) or TRE3G-AR68.FLAG.SPOT (expressing full-length AR68) using the Fugene HD transfection reagent (Promega) and allowed to grow for 48 h. Stable clones were then selected by culturing in the medium containing blasticidin (20 µg/ml) for 2 weeks. AR expression was examined upon induction by doxycycline (1 µg/ml) for 48 h.

Mitochondrial fractionation

Five plates (150 mm diameter) of inducible cells were treated with doxycycline for 2 days and collected in ice-cold phosphate-buffered saline (PBS). After two washes in PBS, cells were resuspended in reticulocyte standard buffer buffer (10 mM HEPES, 10 mM NaCl, 1.5 mM MgCl₂), gently disrupted using a Dounce and tight pestle B with 25 strokes and diluted in cold homogenization buffer (10 mM HEPES, 70 mM sucrose, 210 mM mannitol, 1 mM EDTA). The homogenate was centrifuged twice at 1000g for 10 min at 4°C. The resultant supernatant was further centrifuged at 15 000g for 15 min at 4°C to pellet mitochondrial fractions, which were washed twice in cold buffer before direct use or stored at −80°C.

Statistical analysis

Statistical analyses were performed using Prism version 9 (GraphPad Software, Inc.). One-way ANOVA was used when comparing three conditions or samples (Tukey’s multiple comparisons test) after removing outliers (ROUT, Q = 1%). A P-value of <0.05 was considered significant.

Data availability

All research materials are available upon request through material transfer agreement. Source data are provided with this paper. The mass spectrometry proteomics data are deposited to the MassIVE repository.

Results

Novel isogenic hiPSC models of SBMA

Clinical studies have revealed a general correlation between the CAG repeat number in the AR gene and SBMA severity, but exceptions have also been reported, suggesting that other factors in addition to the CAG repeat size determine age of disease onset as well as the rate of disease progression.^22,23 To explore cell-type-specific molecular mechanisms underlying SBMA, we devised novel isogenic hiPSC models using CRISPR-Cas9 (Fig. 1A, Table 1 and Supplementary Fig. 1). The gene engineering comprised both introduction of expanded polyQ-encoding repeats into wild-type AR in a parental control line (named HT-series) and subtraction or correction of the pathological CAG repeats from mutant AR in a parental disease line (named SB-series), which is critical for corroboration of mutant protein-specific phenotypes. The isogenic disease lines expressing mutant AR with distinct polyQ numbers allow us to assess the polyQ-dependent and length-specific effects on disease pathology. In addition, inclusion of the isogenic AR knockout lines in the models are essential to differentiate gain- versus loss-of-function for mutant AR. Using this strategy, we successfully developed two series of isogenic hiPSC models with high gene-targeting efficiency (3.8–12.8%, Table 1 and Supplementary Figs 2 and 3); no predicted off-target sites were edited in any isogenic lines (Supplementary Fig. 4). In terms of AR expression levels in these hiPSCs, western blot analysis showed that there is no significant difference across different isogenic lines within the same series (Fig. 1B–D).

Figure 1

Isogenic hiPSC models of SBMA. (A) A schematic of the isogenic hiPSC models of SBMA developed using CRISPR. Isogenic models consist of HT- and SB-series, which were derived from the control HT-AR23 and the disease SB-AR68 lines. (B–D) Immunoblotting analysis and quantitation of relative expression levels of AR in isogenic iPSCs compared to that of the loading control by densitometry-based analysis. Anti-AR H280 was used to detect AR on western blot. Data are presented as the mean with 95% CI from three independent experiments.

Table 1

A summary of isogenic hiPSC models of SBMA

HiPSC series	Isogenic lines	CAG or polyQ length in AR	Gene knockout/replacement efficiency (%)b
HT-series	HT-iAR KO	–	62.5% (10/16)
	HT-AR23a	23CAG	–
	HT-iAR40	40Q	9.6% (9/94)
	HT-iAR54	54Q	3.8% (3/79)
	HT-iAR68	68Q	11.4% (4/35)
SB-series	SB-iAR KO	–	37.5% (6/16)
	SB-iAR23	23Q	8.3% (3/36)
	SB-iAR40	40Q	12.8% (11/86)
	SB-iAR54	54Q	3.8% (3/78)
	SB-AR68a	68CAG	–

Parental hiPSC lines.

Gene knockout/replacement efficiency (%) was calculated based on percentage of sequence-confirmed positive clones in the total screened clones.

To differentiate hiPSCs into disease-relevant spinal motor neuron-like cells, we adapted a small-molecule, cocktail-directed approach (Fig. 2A). Notably, the hormones progesterone and corticosterone, which are required for motor neuron differentiation, were withdrawn from the culture medium during the final stage of differentiation (see 'Materials and methods' section), given their possible impacts on AR translocation/transactivity at the given concentrations.²⁴ After a 2-day clearance, cells were treated with either vehicle (EtOH) or ligand (R1881, a synthetic androgen agonist), the latter inducing nuclear translocation of the AR protein. Using this method, we routinely produce >85% motor neuron-like cells, and isogenic lines show similar differentiation efficiency across different experiments (Fig. 2B and C and Supplementary Fig. 5A). In terms of AR expression levels, in contrast to cells at the iPSC stage, differentiated motor neuron-like cells display detectable variations in distinct isogenic lines within the same series (Fig. 2D–F). However, we did not observe aggregation of mutant AR in the differentiated motor neuron-like cells, consistent with other reports.^17,25,26 Regardless of AR expression levels, isogenic disease motor neurons have significantly increased levels of cleaved caspase-3 fragments as compared with isogenic control cells, whether ligand is present or not (Supplementary Fig. 5B). Similar phenotypes are also observed in isogenic AR knockout cells. These data suggest that both loss and gain-of-function for mutant AR possibly contribute to ligand-independent activation of caspase-3-directed apoptosis in SBMA motor neurons.

Figure 2

Motor neuron differentiation of the isogenic hiPSC models of SBMA. (A) A schematic of the adapted approach for motor neuron differentiation. Chemical details are listed in the 'Materials and methods' section. (B) Motor neuron cell identity analysed by immunofluorescence staining of a neuronal marker TUBB3 and two spinal motor neuron markers ChAT and Hb9 . Scale bar = 20 μm. (C) Quantitation of motor neuron differentiation efficiency based on the percentage of ChAT⁺/TUBB3⁺/Hb9⁺ cells in the whole cell population. Twenty images for each isogenic line were randomly selected and subjected to quantitation. Data are presented as the mean with 95% CI from three independent experiments. (D–F) Immunoblotting analysis and quantitation of relative expression levels of AR in motor neuron-like cells compared to that of the loading control by densitometry-based analysis. Anti-AR H280 was used to detect AR on western blot. Data are presented as the mean with 95% CI from three independent experiments. *A non-specific band in western blot analysis; ^$a band reacted with the anti-AR H280 antibody, possibly an AR variant with partially truncated ligand-binding domain expressed in motor neuron-like cells. P-value is shown.

Ligand-free mutant AR leads to mitochondrial impairments in SBMA neurites due to gain-of-function mechanisms

We examined whether these isogenic hiPSC models could recapitulate SBMA phenotypes, such as mitochondrial impairments, upon motor neuron differentiation. First, we measured the mitochondrial membrane potential (Δψ_m) using the cell-permeant, red-orange fluorescent cationic dye TMRE that labels polarized mitochondria with an intact Δψ_m. Our data demonstrate that, when introducing a pathological polyQ number into the wild-type AR, significant mitochondrial depolarization occurs in neurites of the isogenic disease motor neurons compared with parental control cells (Fig. 3A and C). When correcting an expanded polyQ length in the mutant AR, mitochondrial Δψ_m is largely restored in the isogenic control motor neurons compared with parental disease cells (Fig. 3E and G). Interestingly, isogenic AR knockout motor neurons show no significant changes in the maintenance of mitochondrial Δψ_m compared with isogenic control cells (Fig. 3B, D, F and H), implying that aberrant mitochondrial depolarization in disease neurites is possibly due to a gain-of-function from mutant AR. Moreover, this mitochondrial depolarization appears in the absence of ligand and becomes worse upon ligand treatment, suggesting that ligand-free mutant AR is capable of dissipating mitochondrial Δψ_m, whereas the ligand-bound form of the mutant protein aggravates such cytotoxicity in SBMA motor neurons.

Figure 3

Aberrant mitochondrial depolarization in neurites of SBMA motor neuron-like cells. (A and B) and (E and F) Representative images of mitochondrial Δψ_m in neurites of motor neuron-like cells of HT- and SB-series, respectively. MitoTracker stained mitochondria and TMRE labelled polarized mitochondria. Scale bar = 5 μm. (C and D) and (G and H) Quantitation of the relative mitochondria Δψ_m. At least 20 images (>30 frames) for each isogenic line in every experiment were randomly selected and subjected to quantitation. Data are presented as the mean with 95% CI from three independent experiments. P-values are shown.

Mitochondria fulfil a number of crucial cellular functions, including energy production in the form of ATP through oxidative phosphorylation (OXPHOS) and generation of ROS; the latter are by-products of OXPHOS under physiological conditions and are usually tightly regulated by effective cellular antioxidant systems. Yet, an imbalance of production and elimination of ROS can damage various cellular components, resulting in oxidative stress. We assessed mitochondrial oxidative stress using the cell-permeant fluorogenic probe CellRox Green, which exhibits green fluorescence upon oxidation by ROS and subsequently binds to mitochondrial DNA. Our data show that upon introducing a pathological polyQ number into the wild-type AR, a remarkably increased level of mitochondrial oxidative stress occurs in neurites of the isogenic disease motor neurons compared with parental control cells (Fig. 4A and C); after correcting an expanded polyQ length in the mutant AR, mitochondrial ROS levels are greatly reduced in the isogenic control motor neurons compared with parental disease cells (Fig. 4E and G). The isogenic AR knockout motor neurons show no significant differences in mitochondrial ROS levels compared with the isogenic control cells (Fig. 4B, D, F and H), indicating that oxidative stress induced by a high level of mitochondrial ROS in disease neurites is possibly due to a gain-of-function from mutant AR. Furthermore, such oxidative stress appears in the absence of a ligand and becomes worse upon ligand treatment, suggesting that ligand-free mutant AR is capable of inducing mitochondrial oxidative stress, whereas the ligand-bound form of the mutant protein amplifies such cytotoxicity in SBMA motor neurons.

Figure 4

Increased mitochondrial oxidative stress in neurites of SBMA motor neuron-like cells. (A and B) and (E and F) Representative images of mitochondrial oxidative stress in neurites of motor neuron-like cells of HT- and SB-series, respectively. MitoTracker stained mitochondria and CellRox labelled the ROS-oxidized, mitochondrial DNA-bound form of the dye. Scale bar = 5 µm. (C and D) and (G and H) Quantitation of the relative ROS levels. At least 20 images (>30 frames) for each isogenic line in every experiment were randomly selected and subjected to quantitation. Data are presented as the mean with 95% CI from three independent experiments. P-values are shown.

To obtain more evidence for ligand-free mutant AR-induced cytotoxicity, we examined the ultrastructure of mitochondria in neurites of differentiated motor neuron-like cells in the absence of ligand using negative staining and electron microscopy. Our data indicate that, upon introducing a pathological polyQ number into the wild-type AR, abnormal mitochondrial ultrastructure, including significantly shortened cristae, is observed in neurites of the isogenic disease motor neurons as compared with parental control cells (Fig. 5A and C). Upon correcting an expanded polyQ length in the mutant AR, the shortening phenotype of mitochondrial cristae is markedly reversed in the isogenic control motor neurons compared with parental disease cells (Fig. 5E and G). The isogenic AR knockout motor neurons show a similar pattern of normal cristae architecture to mitochondria in isogenic control cells (Fig. 5B, D, F and H), suggesting that the disorganization of mitochondrial cristae in disease neurites is possibly due to a gain-of-function from ligand-free mutant AR.

Figure 5

Altered cristae architecture in neurites of SBMA motor neuron-like cells. (A and B) and (E and F) Representative images of mitochondrial ultrastructure in neurites of motor neuron-like cells of HT- and SB-series, respectively. Scale bar = 200 nm. (C and D) and (G and H) Quantitation of cristae numbers per unit length (µm) of mitochondria. At least 50 images (>80 mitochondria) for each isogenic line were randomly selected and subjected to quantitation. Data are presented as the mean with 95% CI from one experiment. P-values are shown.

The phenotypes described above, including aberrant mitochondrial depolarization, elevated mitochondrial oxidative stress and altered cristae architecture, are observed in isogenic disease motor neurons in both series, providing convincing evidence that these phenotypes are mutant AR-specific and motor neuron-related. Furthermore, most of these phenotypes display a polyQ size-dependent severity. However, some phenotypic discrepancies, possibly due to different genetic backgrounds in the two series, were also observed in our experiments (Supplementary Fig. 9B and C). For instance, higher levels of mitochondrial ROS and greater numbers of long cristae occur in the isogenic control cells of the SB-series (SB-iAR23) as compared to the HT-series (HT-AR23). SB-iAR23 neurons also fail to induce mitochondrial hyperpolarization upon ligand treatment as HT-AR23 neurons do (Fig. 3C and D versus 3G and H).²⁷ Moreover, the HT-series neurons display a significant decrease in mitochondrial size upon introducing a pathological polyQ number into the wild-type AR, while correcting the mutation; no such phenomenon was observed in the SB-series neurons (Supplementary Fig. 6). Together, it is indicative of other mutations related to mitochondrial function existing in the parental disease line of the SB-series (SB-AR68), which makes it difficult to precisely examine the impact of pathological polyQ numbers on the severity of the studied mitochondrial abnormalities when only using one series. Therefore, both series of isogenic lines, including introducing mutations to control cells and correcting mutations in disease cells, are critical for corroboration of mutant protein-specific phenotypes in hiPSC disease modelling.

Compromised mitochondrial respiration in SBMA motor neuron-like cells

Inadequate respiratory capacity in the face of mitochondrial dysfunction contributes to neuronal susceptibility to cellular stress.²⁸ We assessed mitochondrial respiration using the Seahorse XF Cell Mito Stress assay. By measuring the OCR after treating cells with distinct mitochondrial modulators, key parameters of mitochondrial respiration, including basal respiration, ATP-linked respiration and spare respiratory capacity, were compared between isogenic control and disease motor neuron-like cells. Our data demonstrate a significant reduction in oxygen consumption rate as well as other key parameters in differentiated cells after introducing a mutation into wild-type AR, whereas these reductions are suppressed by correcting the expanded polyQ in mutant AR (Fig. 6). The compromised mitochondrial respiration in isogenic disease cells appears in the absence of ligand. Surprisingly, there is no supplementary aggravation upon ligand treatment, indicating that mutant AR impedes ATP production and undermines mitochondrial respiration in SBMA motor neurons in a ligand-independent fashion.

Figure 6

Compromised mitochondrial respiration in SBMA motor neuron-like cells. (A and B) The respiratory phenotypes were defined in motor neuron-like cells of HT- and SB-series, respectively, using the Seahorse XF Mito Stress Assay. Mitochondrial modulators, including oligomycine, FCCP, rotenone and antimycin A, were sequentially injected at specific times during the Seahorse XF Cell Mito Stress Assay according to the manufacturer’s instructions. Oxygen consumption rate (OCR) was calculated and normalized by total protein in the samples. Three key parameters in mitochondrial respiration, including (C and D) basal respiration, (E and F) ATP-linked respiration and (G and H) spare respiratory capacity, in the isogenic control and disease cells were analysed. Data are representative of three independent experiments with at least three replicates per group, and are presented as the mean with 95% CI. P-values are shown.

Preferential interaction between ligand-free mutant AR and F-ATP synthase impairs mitochondrial respiration

AR localizes to mitochondria in muscle of SBMA patients as well as in a mouse motor neuron-derived MN-1 cell line and differentiated pheochromocytoma-derived PC12 cells,^10,11 possibly driven by an N-terminal mitochondrial targeting sequence.¹² Both wild-type and mutant AR interact in vitro with a mitochondrial protein, cytochrome c oxidase subunit Vb (COXVb), which is specifically sequestered by mutant AR aggregates upon ligand treatment.⁹ We surmised that mutant AR may localize to mitochondria and interact with other mitochondrial proteins, leading to ligand-free mutant AR-directed mitochondrial dysfunction in SBMA motor neurons.

We carried out co-IP using an anti-AR antibody (N20) that recognizes the N-terminus of the protein, followed by label-free mass spectrometry and GO enrichment analyses. Among the identified interacting proteins that preferentially interact with mutant AR compared to its wild-type counterpart, mitochondrial function-related proteins, including several subunits of F-ATP synthase, are overrepresented in the isogenic disease motor neurons in the absence of ligand (Supplementary Fig. 7A and B). Furthermore, we used another anti-AR monoclonal antibody to perform co-IP followed by western blot analysis, confirming that subunit D (ATP5H) in the F₀ sector of F-ATP synthase binds more robustly to ligand-free mutant AR over its wild-type counterpart (Supplementary Fig. 7C). Together, our data suggest that aberrant interaction between mutant AR and F-ATP synthase may account for the compromised mitochondrial respiration in SBMA motor neurons.

We sought to further confirm our finding of an aberrant interaction between mutant AR and F-ATP synthase. Given that there is no robust anti-AR antibody available for immunofluorescence staining that is capable of detecting the protein at low abundance in motor neurons, plus practical difficulties in performing mitochondrial fractionation of hiPSC-derived motor neuron-like cells, we developed two additional cell models. These included a FLAG.AR isogenic hiPSC model and an inducible cell model, with both expressing the 3×FLAG.SPOT-tagged wild-type or mutant AR (Supplementary Figs 8 and 10 and 'Materials and methods' section).

After motor neuron differentiation of the FLAG.AR isogenic hiPSC lines, we performed co-IP using FLAG magnetic beads followed by western blot analysis as well as a PLA. The co-IP data show that ligand-free FLAG-tagged mutant AR precipitates two subunits of F-ATP synthase, including ATP5H in the F₀ sector and subunit beta (ATP5B) in the F₁ sector, with much stronger binding to the complex as compared to wild-type AR (Fig. 7A and B). Furthermore, the PLA data show that both wild-type and mutant AR localize to mitochondria and interact with subunit alpha (ATP5A1) in the F₁ sector in neurites of differentiated motor neurons (Fig. 7C and D). Although the genetic backgrounds have significant impact on quantitation of the PLA signals (Supplementary Fig. 9E and F), ligand-free mutant AR in both series exhibits a much stronger interaction to this binding partner as compared to its wild-type counterpart and with the ligand-bound mutant protein (Fig. 7C–H), suggesting a favoured interaction between ligand-free mutant AR and multiple subunits of F-ATP synthase in SBMA motor neurons. In addition, PLA signals were notably decreased upon ligand treatment versus basal conditions, for both wild-type and mutant AR (Fig. 7G and H). These data imply two partitions of mitochondria-localized AR protein: one portion is loosely attached to the outer mitochondrial membrane, which translocates into the nucleus upon ligand binding, whereas the other portion remains within the organelle no matter whether ligand is present. Because ligand treatment fails to aggravate the compromise of mitochondrial respiration in disease motor neurons (Fig. 6), it supports the notion that, regardless of mutant AR translocation upon ligand binding, the mitochondrial distribution of the mutant protein provides temporal and spatial accommodation for its preferential interaction with F-ATP synthase, which impairs the function of this complex and leads to mitochondrial dysfunction.

Figure 7

Aberrant interaction between mutant AR and F-ATP synthase in SBMA motor neuron-like cells. (A and B) Co-IP followed by immunoblotting analysis shows the preferential interaction of two subunits of F-ATP synthase, ATP5B and ATP5H, with FLAG-tagged mutant AR in motor neuron-like cells of HT- and SB-series, respectively, compared to wild-type AR. (C, E and G) and (D, F and H) Representative images of the interaction between AR and ATP5A1 in motor neuron-like cells of HT- and SB-series, respectively, and quantitation of the PLA signals. Scale bar = 5 µm. At least 20 images (>400 PLA puncta) for each isogenic line in every experiment were randomly selected and subjected to quantitation. Data are representative of three independent experiments and are presented as the mean with 95% CI. P-values are shown.

In addition to the disease-relevant motor neurons, we obtained similar results using the inducible cell model. Upon doxycycline induction, U2OS cells stably express the C-terminal SPOT.3×FLAG-tagged wild-type or mutant AR, a sizable fraction of which is mitochondria-localized (Supplementary Fig. 10B and C). After co-IP from the mitochondrial fractions, ATP5B preferentially interacts with mutant AR versus its wild-type counterpart (Supplementary Fig. 10D). Furthermore, the PLA data show that both wild-type and mutant AR localize to mitochondria and interact with ATP5A1 (Supplementary Fig. 10E). Together, these findings indicate that aberrant interactions between mitochondria-localized mutant AR and F-ATP synthase occur not only in disease motor neurons, but also in the inducible cell model, and such interaction is not subject to the position of 3×FLAG.SPOT tag, no matter whether the tag is at the N- or C-terminus of the target protein. Recent studies suggest that F-ATP synthase is a core component of the mitochondrial permeability transition pore (mPTP),^29,30 which is a key effector of cell death. Thus, it remains possible that the interaction between mutant AR and F-ATP synthase promotes the long-lasting opening of the mPTP.

Discussion

The discovery of the causative triplet repeat expansion mutation of SBMA was made 30 years ago. Since then, diverse models ranging from in vitro settings to cell and animal models have been studied extensively, identifying various genetic modifiers, post-translational modifications and aberrant protein interactions as potential disease aetiologies. Multiple clinical trials have also been conducted in order to understand molecular and cellular mechanisms underlying SBMA and to search for and test treatments for this devastating disease. Yet, there is still no effective therapy to stop or reverse the course of SBMA.

Studies in heterozygous and homozygous females emphasize the requirement of androgens in SBMA manifestations^31,32; however, removing hormones from male patients has little effect on disease progression in post-symptomatic patients. Even with a pre-symptomatic, anti-androgen medication for many years, a male-to-female transgender patient still developed full SBMA manifestations and her disease progression was not delayed relative to her brother.³³ Thus, we must consider whether mutant AR itself, without ligand binding, is adequate to induce cytotoxicity in SBMA patients. The findings we report here provide substantial evidence, for the first time, that ligand-free mutant AR is sufficient to lead to disease pathology (i.e. mitochondrial dysfunction) in hiPSC-derived SBMA motor neurons.

Previously, a large body of evidence has implicated mitochondrial dysfunction in SBMA pathogenesis. In these studies, mutant AR elicited toxic effects on mitochondria in both disease-related and unrelated cell types. This is also the case for Huntington's disease, another CAG-repeat expansion disorder, implying a general toxic nature of the pathological polyQ-containing mutant proteins without cell-type selectivity.³⁴ Instead, cell-type-specific characteristics might render distinct neuronal populations vulnerable against polyQ-induced toxicity, which lead to disease-related, context-dependent neurodegeneration. Thus, using isogenic hiPSC modelling to elucidate disease-relevant, cell-type-specific molecular mechanisms is particularly urgent for our understanding of disease aetiology and will effectively direct our efforts for therapeutic design. Together with other recent findings made in hiPSC-derived motor neurons,^35,36 our data unravel different molecular mechanisms underpinning the pathologic polyQ-dependent, mitochondria-targeted cytotoxicity in SBMA motor neurons, which involve ligand-free (specifically, mitochondria-localized) and ligand-bound (nuclear-localized) forms of mutant AR. Furthermore, cytoplasmic compartmentalization seems to modulate the traits of ligand-free mutant AR, as the mitochondria-localized form was toxic, whereas, on the contrary, the cytoplasm-localized form was protective in mouse motor neurons in primary culture.³⁷ These observations introduce a new and exciting element in the quest for understanding mutant AR-induced cytotoxicity in the context of motor neuron-related SBMA pathogenesis.

Since the first observation of mitochondrial abnormalities (i.e. altered mitochondria distribution) in NSC34 motor neuron-like cells,³⁸ an intriguing question has arisen: is mitochondrial dysfunction a consequence of motor neuron degeneration or a contributor to SBMA manifestations? A prevailing view regarding the significance of mitochondrial impairments in neurodegeneration underlying many neurodegenerative disorders is that mitochondrial dysfunction represents an early event eliciting the shift from normal physiological towards pathological conditions.⁷ This ‘switch’ can be triggered by occurrence and/or accumulation of pathologic mutant proteins through direct association with mitochondria and/or interaction with other mitochondrial proteins. For instance, mutant HTT interacts with TIM23 to inhibit mitochondrial protein import in Huntington's disease models.³⁹ In this study, we have demonstrated that mutant AR has mitochondrial localization and interacts with F-ATP synthase in the hiPSC-derived motor neurons, accounting for the compromised mitochondrial respiration and a series of mitochondrial deficits. Our findings elucidate a mechanistic molecular basis for ligand-free mutant AR/F-ATP synthase-involved mitochondrial dysfunction, highlighting a critical driver to motor neuron degeneration in SBMA. However, the exact topology of mutant AR inside mitochondria remains unclear at this time. Because F-ATP synthase is a large complex situated on the inner mitochondrial membrane and matrix, it is possible that mutant AR bypasses the outer mitochondrial membrane to directly interact with the complex, although it remains possible that another mitochondrial protein aids the interaction, bridging the two proteins.

F-ATP synthase is a multi-subunit enzymatic complex that catalyses the conversion of ADP to ATP to produce most of the energy for the cell. F-ATP synthase forms dimers, and these dimer assemblies induce mitochondrial membrane curvature and contribute to the formation and maintenance of proper cristae architecture.⁴⁰ Moreover, F-ATP synthase has recently emerged as a core component of the mPTP, whose acute or prolonged opening orchestrates physiological and pathological roles, respectively. Nevertheless, in severe pathological settings, F-ATP synthase tends to hydrolyse ATP, turning the ‘powerhouse’ into an energy ‘consumer’.⁴¹ Hence, F-ATP synthase has been implicated in several neurodegenerative disorders. Interestingly, some subunits of F-ATP synthase also interact with other pathologic proteins, including beta-amyloid, tau, alpha-synuclein, DJ-1 and C9orf72 repeat-derived poly (GR).⁴² Taking account of the findings amassed to date, the combination of our data using the hiPSC-derived motor neurons and the inducible cell model leads us to propose that the aberrant interaction between ligand-free mutant AR and F-ATP synthase provokes detrimental changes to structure and function of the latter, enhancing the induced cytotoxicity of the mutant protein. This interplay accounts for the compromised mitochondrial respiration and abnormal mitochondrial architecture, leading to increased oxidative stress, dissipation of mitochondrial Δψ_m and activation of apoptosis in SBMA motor neurons, and ultimately, neuronal death.

In addition to defining the mechanisms involved, our findings may provide a new path to design and develop motor neuron-specific therapeutic strategies to rescue the overall health and function of mitochondria in SBMA. Defining the interface of the mutant AR/F-ATP synthase interaction, whether new binding site(s) are involved or the same binding site(s) on mutant AR but with significantly increased affinity as compared to that on wild-type AR and/or addressing helper protein(s) if they exist, are of particular significance, which will require the combined efforts of experimental and computational modelling approaches.

Acknowledgements

We are grateful to Alan Koretsky, Lorna Role, Kenneth Fischbeck and Christopher Ross for support; to Emma Joyner, Ronald Wang, Jake Sun and Aiman Zehra Altaf for technical assistance; to the Protein/Peptide Sequencing Facility and Electron Microscopy Facility at NINDS, iPSC Core at NHLBI, Flow Cytometry Core at NHLBI, and Microscope Facility at Johns Hopkins University School of Medicine for mass spectrometry, electron microscopy, gene knockout in iPSCs, fluorescence-activated cell sorting, and confocal microscopy experiments, respectively; to Christopher Grunseich, Zhong-Wei Du and Su-Chun Zhang for stem cell technique and motor neuron differentiation protocols; and to Pierre A. Coulombe and Ryan Hobbs for advice.

Funding

This research was funded by NINDS Competitive Postdoctoral Fellowship Award/Intergovernmental Personnel Act Agreement, Kennedy’s Disease Association Research Grant and Maryland Stem Cell Research Fund Launch Award (#135408, to X. F.), as well as the Intramural Research Program of the NINDS, National Institutes of Health.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.