Cardiac glycosides restore autophagy flux in an iPSC-derived neuronal model of WDR45 deficiency

Abstract

Beta-Propeller Protein-Associated Neurodegeneration (BPAN) is one of the commonest forms of Neurodegeneration with Brain Iron Accumulation, caused by mutations in the gene encoding the autophagy-related protein, WDR45. The mechanisms linking autophagy, iron overload and neurodegeneration in BPAN are poorly understood and, as a result, there are currently no disease-modifying treatments for this progressive disorder. We have developed a patient-derived, induced pluripotent stem cell (iPSC)-based midbrain dopaminergic neuronal cell model of BPAN (3 patient, 2 age-matched controls and 2 isogenic control lines) which shows defective autophagy and aberrant gene expression in key neurodegenerative, neurodevelopmental and collagen pathways. A high content imaging-based medium-throughput blinded drug screen using the FDA-approved Prestwick library identified 5 cardiac glycosides that both corrected disease-related defective autophagosome formation and restored BPAN-specific gene expression profiles. Our findings have clear translational potential and emphasise the utility of iPSC-based modelling in elucidating disease pathophysiology and identifying targeted therapeutics for early-onset monogenic disorders.

Figure 1.. BPAN and control iPSC line differentiation into mDA neurons

(A) Immunofluorescence analysis of MAP2/TH in patient and control lines at day 65 of differentiation. Mature mDA neurons exhibit typical morphology and express pan-neuronal (MAP2) and mDA-specific (TH) markers. Nuclei were counterstained with DAPI. Scale bar, 200 μm. n=3 biological replicates per line. (B) Quantification of MAP2/TH abundance in control and patient-derived neurons (n=3 biological replicates per line, 3 individual images analysed per replicate). Percentages were calculated after manual counting of cells on ImageJ/Fiji (approximately 500 nuclei counted per image, followed by counting of cells also staining positive for TH and/or MAP2). (C) Chromatograms from WDR45 cDNA sequencing in female BPAN iPSC lines. In Patient 01, the cDNA has 28 additional bp, while for Patient 03, the c.700C>T leading to an early stop codon is retained. (D) Relative Fluorescence Units of Androgen Receptor CAG repeat PCR products from female BPAN lines, in the presence or absence of methylation-sensitive restriction enzyme Hpall. Patient 01 fibroblasts (BUCL01) exhibit a random XCI pattern, with two PCR bands detectable in the presence and absence of Hpall. For patient 03 (535–201), there is practically only one detectable band in fibroblasts (535–201), signifying a skewing of XCI towards the expression of only one allele. For both patients, iPSC lines and derived neurons have very skewed patterns of XCI, with PCR bands practically undetectable in the presence of Hpall. n=1 biological replicate per line. (E) Cropped immunoblot of total WDR45 and beta actin protein expression at Day 65, and quantification of WDR45 relative to the loading control. n=3 biological replicates for each line. Error bars represent the Standard Error of Mean. Statistics were calculated using ANOVA.

Figure 2.. RNASeq at Day 65 of differentiation.

(A) Volcano plots comparing gene expression profiles patient lines versus controls. Overexpressed genes are shown on the right of the X axis while underexpressed genes of the left. (B) Heat maps depicting gene underexpression and overexpression in gradients of green and red, respectively. (C) GO Term and KEGG pathway enrichment analysis depicting intracellular pathways affected in BPAN. Pathways with the most significant Fold Enrichment (Y axis) are shown; differentially expressed gene count is depicted on the X axis. n=3 biological replicates for all lines, analysis based on median TPM values. P-values of <0.05 and fold change of <0.5 or >2 [-11] (Student’s t-test) were set as statistically significant cut-offs. The top 40 genes (as per lowest p-values) are labelled in volcano plots and heat maps. ATG= autophagy-related gene, TPM= transcript per million.

Figure 3.. High content imaging LC3 assay for drug screening.

(A) High Content Imaging immunofluorescence analysis of LC3/ FOXA2 in patient and control ventral midbrain progenitors at day 11 of differentiation. Representative images. Cell density at 15,000 cells/well. n=8 independent differentiations/ biological replicates for each line. (B) Quantification of LC3 puncta/nuclei in control and patient-derived neurons DMSO and Torin 1 treatments were of 3-hour durations. n=8 independent differentiations/ biological replicates for each line, each condition per line tested in technical duplicates, 20 fields imaged per well. For statistical analysis, the Student’s unpaired two tailed t-test was used. Error bars represent the Standard Error of Mean.

Figure 4.. Drug Screening

(A) Screening protocol. Patient 02 cells used for the screen. The process was optimised and, wherever possible, automated to allow large-scale compound testing. Overall, 16 48x 96-well destination plates were required (16 source plates, 80 compounds each, tested in technical triplicates). (B) Source plate layout (left) and a representative heatmap (right) are shown. N and P represent negative and positive controls, respectively. Increasing puncta numbers per nuclei (in hits and positive controls) are depicted in increasingly darker shade of red. (C) Representative images. Immunofluorescence analysis of LC3/ DAPI after treatments with DMSO, compound hit digoxin and Bafilomycin A1. Digoxin increases LC3 puncta production to levels higher than in Bafilomycin-treated cells. Scale bar, 200 μm (D) Distribution of hits with the highest z-scores. There is an even distribution across all 16 source plates. (E) Quantile-Quantile plot for screened compounds. Several compound hits have z-scores and sample quantiles that are much higher than their theoretical values, as well as values from positive controls (in red).

Figure 5.. P62 hit validation assay.

(A) Immunofluorescence analysis of p62/DAPI in Patient 02 Day 11 ventral midbrain progenitors after treatments with Prestwick screen hits with the 26 highest z-scores. Representative images from negative controls (DMSO), positive controls (Torin 1), autophagy inducers and blockers, as well as ‘false hits’ (resulting in reduced cell viability). (B) Quantification of cell viability (fold change of nuclei number per well, compared to DMSO) and p62 degradation (fold change of p62 puncta per nuclei, compared to DMSO; green bars) for each hit, as well as Torin 1 and DMSO. (C) 5 cardiac glycosides significantly degrade p62 when compared to DMSO and are associated with acceptable cell viability. There is no statistical difference in p62 puncta per cell, or nuclei per well, when comparing Torin 1 and the 5 cardiac glycosides. There is also no statistical significance between DMSO-treated and glycoside-treated nuclei per well (statistics not shown). n=3 technical replicates for each compound, n=9 for DMSO and Torin 1. Error bars represent Standard Deviation. Statistics were calculated using ANOVA. Abbreviations: FC = fold change, Nr= number

Figure 6.. Correction of intracellular pathways after compound hit 48-hour treatment.

(A) Volcano plots comparing Patient 02, Patient 02 compound-treated and Patient 02 CRISPR-corrected (CRISPR 01) mDA neuronal gene expression. (B) Intracellular pathways jointly corrected by both WDR45 CRISPR-mediated mutation correction and compound treatments. (C) ATG differential expression and correction after CRISPR-genome editing and compound treatments. ATGs with consistent differential (over- or under) expression after CRISPR correction and compound treatments, when compared to untreated Patient 02 neurons, are marked with *. White= not mapped during RNAseq. (D), (E) ClueGO analysis of GO term enrichment of differentially expressed genes, showing network graphs of differentially expressed genes between Patient 02, CRISPR01, and compound-treated Patient 02 mDA neurons. Cellular components (CC) (C), and molecular functions (MF) (D) are shown. Network graph nodes represent GO terms (the most significant are named) and edges indicate shared genes between GO terms. Functional groups of GO terms are indicated by the same colour. GO functional groups exhibiting statistically significant differences (p< 0.05) are shown. n=3 biological for all lines, analysis based on median TPM values. P-values of <0.05 and fold change of <0.5 or >2 [-1<log2(FC)>1] (Student’s t-test) were set as statistically significant cut-offs. The top 40 genes (as per lowest p-values) are labelled in volcano plots.