Dipeptide repeat proteins inhibit homology-directed DNA double strand break repair in C9ORF72 ALS/FTD

Abstract

Background: The C9ORF72 hexanucleotide repeat expansion is the most common known genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), two fatal age-related neurodegenerative diseases. The C9ORF72 expansion encodes five dipeptide repeat proteins (DPRs) that are produced through a non-canonical translation mechanism. Among the DPRs, proline-arginine (PR), glycine-arginine (GR), and glycine-alanine (GA) are the most neurotoxic and increase the frequency of DNA double strand breaks (DSBs). While the accumulation of these genotoxic lesions is increasingly recognized as a feature of disease, the mechanism(s) of DPR-mediated DNA damage are ill-defined and the effect of DPRs on the efficiency of each DNA DSB repair pathways has not been previously evaluated.

Methods and results: Using DNA DSB repair assays, we evaluated the efficiency of specific repair pathways, and found that PR, GR and GA decrease the efficiency of non-homologous end joining (NHEJ), single strand annealing (SSA), and microhomology-mediated end joining (MMEJ), but not homologous recombination (HR). We found that PR inhibits DNA DSB repair, in part, by binding to the nucleolar protein nucleophosmin (NPM1). Depletion of NPM1 inhibited NHEJ and SSA, suggesting that NPM1 loss-of-function in PR expressing cells leads to impediments of both non-homologous and homology-directed DNA DSB repair pathways. By deleting NPM1 sub-cellular localization signals, we found that PR binds NPM1 regardless of the cellular compartment to which NPM1 was directed. Deletion of the NPM1 acidic loop motif, known to engage other arginine-rich proteins, abrogated PR and NPM1 binding. Using confocal and super-resolution immunofluorescence microscopy, we found that levels of RAD52, a component of the SSA repair machinery, were significantly increased iPSC neurons relative to isogenic controls in which the C9ORF72 expansion had been deleted using CRISPR/Cas9 genome editing. Western analysis of post-mortem brain tissues confirmed that RAD52 immunoreactivity is significantly increased in C9ALS/FTD samples as compared to controls.

Conclusions: Collectively, we characterized the inhibitory effects of DPRs on key DNA DSB repair pathways, identified NPM1 as a facilitator of DNA repair that is inhibited by PR, and revealed deficits in homology-directed DNA DSB repair pathways as a novel feature of C9ORF72-related disease.

Keywords: Amyotrophic lateral sclerosis; CRISPR; DNA damage; DNA double strand break repair; Homology-directed repair; Induced pluripotent stem cells; RAD52; Single-strand annealing.

Fig. 1

Subcellular localization of PR. a) Representative confocal images of U-2 OS cells stably expressing wild-type NPM1 (WT), NPM1 with deletions of the nuclear export (NESΔ) or nuclear localization (NLSΔ); NPM1 proteins were fused to GFP (green). Each NPM1 construct was co-transfected with PR (red) and nuclei stained with DAPI (blue). Automated microscopy and image analysis was used to quantify levels of PR in the nucleus (b) or cytoplasm (c) as a function of area (average intensity). Statistical significance was assessed by one-way ANOVA and post-hoc test between each experimental group; n = 2 biological groups, 9 fields per group, ****P < 0.0001; error bars are SEM, dots represent single cells. d) Representative images of U-2 OS cells stained with DAPI (blue) expressing GFP-NPM1 fusion proteins (green) with deletions of the nuclear localization (NLSΔ) or nucleolar (NuLSΔ) signals to confer cytoplasmic localization. Cells were immunolabeled with an antibody against PR (red). e) Nuclear/Cytoplasmic mean fluorescence (Y-axis) for n > 30 cells per mutant cell line (X-axis), n = 3 biological replicates. *P < 0.05; ***P < 0.0005 student’s t-test; error bars are SEM. f) Cartoon (created in PyMOL) with top and side perspectives of the NPM1 (PDB 4N8M) pentamer

Fig. 2

Efficiency of double strand DNA break repair pathways in response to dipeptide repeat proteins. a-d) Relative repair efficiency (Y-axis) as determined by the percentage of GFP positive cells in cultures transfected with DPR expression plasmids or an empty vector (set to 100% efficiency). Four reporter cell lines were used to assess the efficiency of (a) homologous recombination (HR), (b) non-homologous end joining (NHEJ), (c) microhomology mediated end joining (MMEJ), and (d) single strand annealing (SSA). Statistical significance was assessed by one-way ANOVA and post-hoc test between each experimental group and the control group (vector); n = 6 biological replicates, 100,000 cells/replicate were evaluated; error bars are SEM, *p < 0.05, ***p < 0.0005, ****p < 0.0001. Inset numbers are the mean difference between groups. E-H) Representative fluorescence activated cell sorting plots of transfected NHEJ and SSA reporter cells using GFP fluorescence (Y-axis) and side scatter (X-axis); the number of GFP positive cells is represented as a percentage of parent gating

Fig. 3

DNA repair efficiency in response to the manipulation of nucleophosmin levels. a-b) Repair efficiency (Y-axis) as determined by the percentage of GFP positive cells in cell cultures transfected with proline-arginine (PR), glycine-arginine (GR), glycine-alanine (GA) and nucleophosmin siRNA, or control siRNA. Two reporter cell lines were used to assess the efficiency of (a) non-homologous end joining (NHEJ) and (b) single strand annealing (SSA). Statistical significance was assessed by one-way ANOVA and post-hoc test between each experimental group and the control group (control siRNA); n = 6 biological replicates, 100,000 cells/replicate were evaluated; error bars are SEM, *p < 0.05, ***p < 0.0005, ****p < 0.0001. (c-f) Representative fluorescence activated cell sorting plots of transfected NHEJ and SSA reporter cells using GFP fluorescence (Y-axis) and side scatter (X-axis); the number of GFP positive cells is represented as a percentage of parent gating

Fig. 4

Super-resolution Stochastic Optical Reconstruction Microscopy (STORM) shows nuclear co-localization of nucleophosmin and phospho-RAD52. a) Representative images of NPM1 immunostaining (green) in the nuclei (white oval traces in top panels) of cells treated with etoposide or vehicle control. Red boxes indicate area of increased magnification in bottom panels; yellow arrows indicate nucleoli; white arrows indicate NPM1 clusters. b-c) Quantification of NPM1 clustering within the nucleoplasm of 3 cells with and without etoposide treatment. ****P < 0.0001, as determined by unpaired student’s t-test, error bars are SEM. d) Representative super-resolution analysis of U2-OS cells treated with etoposide to induce DNA DSBs or vehicle control then stained with antibodies against NPM1 and pRAD52. Colored heat-map where red indicates positive spatial overlap of NPM1 and pRAD52 (correlation coefficient r = 1) and blue indicates negative correlation (r = − 1). E) Numerical quantification of NPM1 and pRAD52 co-localization in the nucleus, n = 10 cells for each condition. Significance was assessed by unpaired student’s t-test (**P < 0.01); error bars are SEM

Fig. 5

Expression of activated and total RAD52 in C9ALS/FTD neurons. a) Quantification of pRAD52 mean fluorescence in neurons using automated image analysis software (Fiji/Image J) for three C9ALS iPSC lines normalized to isogenic lines; each data point represents one cell. b) Representative confocal immunofluorescence images of iPSC motor neuron cultures stained with an antibody against phospho-RAD52 (pRAD52) (green) and counter-stained with DAPI (blue) scale bars are 100 μm. c) Quantification of total RAD52 (tRAD52) mean fluorescence normalized to isogenic line. d) Representative images of iPSC motor neuron cultures immuno-stained with an antibody against RAD52 (tRAD52) (green) and DAPI (blue). n = 3 biological replicates, 5 fields per replicate, error bars are SEM; ****P < 0.0001, as determined by unpaired student’s t test

Fig. 6

Quantification of total RAD52 in human brain samples. a) Western blot quantification of total RAD52 from unaffected controls (CTL), C9ORF72-related ALS (C9ALS) and sporadic ALS (sALS) in three different brain regions: Occipital cortex (OC) Cerebellum (CB) and Motor cortex (M1). Comparisons between diagnosis groups were performed by mixed effect analyses utilizing data from all three brain regions and accounting for both the between-region differences and within-person correlation; n = 6 per diagnosis group, 3 measurements per person, one from each region. b) Representative western for total RAD52 and beta-actin